Integration Report — n8n

Step 0

Licence conflict check

Run before any integration. n8n’s source-visible product shape is compatible with the authoritative embedded-product grant; the current blocker is only the absent purchase/grant record.

Sub-package scan: 69 n8n-sustainable-use workspaces, 15 MIT, 2 ISC, 1 Apache-2.0. Recommended attachment points avoid the permissively published workspaces.

Verdict

Important unresolved gap: the catalog's rate-limiters do not expose a shared-state/store port, so this report does not claim they solve n8n multi-main limiter state.

Step 5A — the section that sells

Composite components built for n8n

Four host-specific capabilities that exist only when catalog parts are combined with n8n.

COMPOSITE #1 — Execution Evidence Capsule
  Composition:   CMPSBL BLD-ENG-009 "Tamper-Evident Chain" (createAuditChain) + CMPSBL BLD-ENG-105 "Forensic Replay" (AuditEvent, ReplaySession, ReplayResult, ForensicReplayEngine) + CMPSBL STIER-S-81 "Compliance Attestation Generator" (EvidenceItem, Attestation, generateAttestation, batchAttest) + host packages/cli/src/eventbus/**
  Problem:       packages/cli/src/eventbus/message-event-bus-writer/message-event-bus-log-writer-worker.ts:89 appends JSON lines directly; no cryptographic event-chain linkage was found in the observed event-bus implementation. Supply-chain attestations do not prove what an individual workflow execution did.
  Behaviour:     Selected execution/audit events are canonicalized into an append-only hash chain, sliced into a read-only incident replay, and emitted as an evidence attestation. Existing event delivery and execution storage remain authoritative.
  New because:   The chain alone only detects history alteration; replay alone only reconstructs; attestation alone only packages evidence. Together with n8n's existing event stream they create a per-execution evidence capsule that neither side has by itself.
  Attachment:    packages/cli/src/eventbus/message-event-bus-writer/** + event-message-classes/**; optional read-only execution/report endpoint later
  Untouched:     Workflow execution semantics, event-bus destinations, Sentry/error reporting, SBOM/VEX pipeline and existing database execution records.
  Effort:        ~3–6 engineer-days (basis: 67 + 103 + 57 entry LOC, canonicalization/hash/replay/attestation concerns, persistence and endpoint tests).
  Verified:      C:ok I:ok B:— T:— S:scan
  Data path:     execution/audit event -> canonical evidence -> hash chain -> incident slice -> attestation
                         |                    |                |
                         +---- existing event bus/database ----+

COMPOSITE #2 — Provider Failure Cell
  Composition:   CMPSBL SUBP-06 "BULKHEAD" + CMPSBL BLD-ENG-266 "Retry Budget" + CMPSBL BLD-ENG-303 "retry-with-backoff" + host retryability classifier / AI service retry / external-secrets retry paths
  Problem:       packages/cli/src/utils/ai-service-retry.ts:27-29 explicitly says the timed-out underlying SDK request cannot be aborted; packages/cli/src/modules/external-secrets.ee/retry-manager.service.ts:62-91 recursively schedules retries with increasing attempt count and no observed attempt ceiling. No generic per-dependency concurrency compartment was found in the observed outbound transport.
  Behaviour:     For n8n-owned external dependencies only, each provider gets a bounded concurrency compartment; retries use jitter and a refilling retry budget so failures cannot multiply work without bound. Domain-specific retryability classification remains host-owned.
  New because:   Backoff controls timing, the retry budget controls aggregate retry pressure, and the bulkhead controls concurrent occupancy. Only the composition turns one flaky dependency into a bounded failure domain.
  Attachment:    packages/cli/src/utils/ai-service-retry.ts; packages/cli/src/modules/external-secrets.ee/retry-manager.service.ts; selected internal outbound call sites
  Untouched:     User-selected workflow providers, generic HTTP Request node semantics, existing SSRF checks, existing circuit breaker, Redis queueing and workflow retryOnFail contract.
  Effort:        ~3–6 engineer-days (basis: 539 + 56 + 55 entry LOC, dependency partitioning, queue/reject policy, retry budget, jitter, metrics and parity tests).
  Verified:      C:ok I:ok B:— T:— S:scan
  Data path:     n8n-owned call -> dependency bulkhead -> retryability -> budgeted jittered retry -> existing transport
                                      |                  |
                                      +-- saturation ----+--> shed/surface

COMPOSITE #3 — Causal Run Ledger
  Composition:   CMPSBL BLD-ENG-034 "Causal Event Propagation" + CMPSBL BLD-ENG-009 "Tamper-Evident Chain" + host AbstractEventMessage / workflow / node / execution / queue / runner events
  Problem:       packages/cli/src/eventbus/event-message-classes/abstract-event-message.ts:84-92 defines id, timestamp, event name, message and payload but no explicit parent/causal field. Chronology alone can be ambiguous across queue, runner and node boundaries.
  Behaviour:     Deterministic relationships add parent/causal edges to selected events; the resulting chain is hash-linked so an incident view can answer not only what happened next, but what caused what.
  New because:   Causal propagation creates the graph, while the tamper chain protects the graph's recorded history. n8n supplies the real execution events that make the generic primitives useful.
  Attachment:    packages/cli/src/eventbus/event-message-classes/** and the event emission adapters where executionId/node/run relationships are already known
  Untouched:     Existing event ids, payload contracts for destinations, workflow scheduler/executor and database records.
  Effort:        ~2–4 engineer-days (basis: 106 + 67 entry LOC, relationship mapping, canonical hashing and destination-compatibility tests).
  Verified:      C:ok I:ok B:— T:— S:scan
  Data path:     workflow -> node -> queue -> runner events
                    |        |        |
                    +---- causal edges ----> hash-linked run ledger

COMPOSITE #4 — Boundary Regression Replay Pack
  Composition:   CMPSBL BLD-ENG-248 "Adversarial Input Generator" + CMPSBL BLD-ENG-105 "Forensic Replay" + host body-parser / SSRF / endpoint-rate-limit test harnesses
  Problem:       packages/cli/src/middlewares/__tests__/body-parser.test.ts:16-58 and packages/@n8n/backend-network/src/ssrf/__tests__/ssrf-protection.service.test.ts:53-110 show strong boundary-specific tests, while packages/cli/src/services/rate-limit.service.ts:31-74 defines another security boundary; no shared repository-wide adversarial corpus was found across them.
  Behaviour:     Generates deterministic security-boundary fixtures, captures the exact observed result, and promotes any interesting failure into a replayable regression case. Production request handling is not changed.
  New because:   A finite adversarial catalog becomes operationally valuable only when a failure can be captured and replayed through n8n's real parsers/guards.
  Attachment:    tests around packages/cli/src/middlewares/body-parser.ts, packages/@n8n/backend-network/src/ssrf/ssrf-protection.service.ts and packages/cli/src/services/rate-limit.service.ts
  Untouched:     Production parsers, SSRF allow/block rules, limiter thresholds and user workflow semantics.
  Effort:        ~2–4 engineer-days (basis: 33 + 103 entry LOC, adapter/corpus/result-normalization/replay concerns and CI fixture wiring).
  Verified:      C:ok I:ok B:— T:— S:scan
  Data path:     adversarial corpus -> existing boundary under test -> captured result -> forensic replay regression

Composite ledger

Build this composite first: Execution Evidence Capsule. It adds a new evidence property without changing workflow execution semantics.

Step 5B — new single-component designs

New components designed for this product

Five concrete product/engineering designs derived from the retained rows and host evidence.

DESIGN #1 — Enterprise Execution Receipt
  Built from:     CMPSBL BLD-ENG-009 "Tamper-Evident Chain" (createAuditChain) + CMPSBL STIER-S-81 "Compliance Attestation Generator" (EvidenceItem, Attestation, generateAttestation, batchAttest)
  Plus host:      packages/cli/src/eventbus/** execution/audit events and existing execution identifiers
  What it is:     A read-only receipt for a workflow execution containing selected event evidence, chain verification status and a portable attestation payload.
  New because:    n8n stores executions and emits events; the library adds tamper-evident packaging rather than another execution store.
  Unlocks:        Enterprise incident handoff, support evidence, audit export and a stronger trust story without claiming compliance certification.
  Evidence path:  executionId -> selected EventMessage records -> canonical evidence -> createAuditChain -> generateAttestation -> receipt/report
  Effort:         ~3–5 engineer-days (basis: 67 + 57 entry LOC + host event schema/endpoint/tests).
  Constraints:    BLD-ENG-009 Verified: C:ok I:ok B:exec T:— S:scan; mode: direct; ports: none stated. STIER-S-81 Verified: C:ok I:ok B:load T:— S:scan; mode: direct; ports: none stated. T:— = Not claimed. S:scan is the author's self-scan, not an independent audit. Purchase/grant + attribution required before integration.
  Confidence:     high — host evidence seam is explicit and TypeScript-direct, but source review is still required post-purchase.

DESIGN #2 — Provider Isolation Profile
  Built from:     CMPSBL SUBP-06 "BULKHEAD" (BulkheadOptions, BulkheadRunOptions, Bulkhead, createBulkhead, BulkheadSubject, bulkheadContract, BulkheadRejection, BulkheadPolicy, RejectionReason +1 more)
  Plus host:      packages/@n8n/backend-network/src/http/retryability.ts; packages/@n8n/backend-network/src/http/undici/transport.ts; n8n-owned dependency names
  What it is:     A small registry that assigns concurrency/queue/reject policy per internal dependency, deliberately excluding arbitrary user HTTP destinations at first.
  New because:    n8n has circuit breaking and retry classification but no generic per-dependency bulkhead was found in the observed transport.
  Unlocks:        One hung provider cannot occupy the whole process; saturation becomes explicit and measurable.
  Evidence path:  dependency key -> createBulkhead -> existing call -> state snapshot/telemetry
  Effort:         ~2–4 engineer-days integration (basis: 539 entry LOC + host keying/policy/telemetry tests; acquisition-equivalent size is in the kit's 1–3 engineer-week band).
  Constraints:    SUBP-06 Verified: C:ok I:ok B:load T:— S:scan; mode: direct; ports: none stated; constraints: none. T:— = Not claimed. Queue/reject behavior must be introduced behind flags and measured.
  Confidence:     high for n8n-owned dependencies; low for generic workflow traffic until semantics are proven.

DESIGN #3 — Retry Policy Registry
  Built from:     CMPSBL BLD-ENG-303 "retry-with-backoff" (RetryOptions, retry) + CMPSBL BLD-ENG-266 "Retry Budget" (tryRetry, remainingBudget, configureBudget, getBudgetStats, resetBudget)
  Plus host:      packages/@n8n/backend-network/src/http/retryability.ts and selected existing retry call sites
  What it is:     Shared retry policies keyed by dependency/job class: same jitter/backoff primitive, same budget accounting, host-owned retryability predicate.
  New because:    It does not invent retry semantics; it removes repeated mechanics while preserving domain decisions in n8n.
  Unlocks:        Consistent cancellation/jitter, retry-pressure metrics, easier policy review and fewer copy-pasted loops.
  Evidence path:  host failure -> retryability classifier -> tryRetry/remainingBudget -> retry(backoff/jitter) -> existing operation
  Effort:         ~2–4 engineer-days integration (basis: 55 + 56 entry LOC + migration/parity tests across selected helpers).
  Constraints:    BLD-ENG-303 Verified: C:ok I:ok B:— T:— S:scan; mode: direct; ports: none stated. BLD-ENG-266 Verified: C:ok I:ok B:— T:— S:scan; mode: direct; ports: none stated. B:— = Not claimed; T:— = Not claimed. Do not migrate workflow retryOnFail until user-visible parity is proven.
  Confidence:     medium — useful, but much of the concept already exists in host-specific form.

DESIGN #4 — Causal Incident Explorer
  Built from:     CMPSBL BLD-ENG-034 "Causal Event Propagation" (CausalEvent, PropagationRule, CausalChain, addRule, clearRules, propagate, getChain, getAllChains, getStats) + CMPSBL BLD-ENG-105 "Forensic Replay" (AuditEvent, ReplaySession, ReplayResult, ForensicReplayEngine)
  Plus host:      packages/cli/src/eventbus/event-message-classes/** plus executionId/workflowId/node/queue/runner context
  What it is:     A support/debug view that reconstructs a single execution as a cause-effect graph and lets an investigator replay the recorded event slice without re-running side effects.
  New because:    Existing execution history is chronological/data-centric; this design adds explicit cause edges plus a forensic session abstraction.
  Unlocks:        Faster root-cause analysis for distributed worker/queue/runner incidents.
  Evidence path:  event stream -> deterministic parent rules -> causal chain -> read-only ReplaySession -> incident graph
  Effort:         ~3–6 engineer-days (basis: 106 + 103 entry LOC + event relationship mapping/UI or export adapter/tests).
  Constraints:    BLD-ENG-034 Verified: C:ok I:ok B:— T:— S:scan; mode: direct; ports: none stated. BLD-ENG-105 Verified: C:ok I:ok B:— T:— S:scan; mode: direct; ports: none stated. B:— = Not claimed; T:— = Not claimed. Causal claims must be limited to relationships the host can prove.
  Confidence:     medium — valuable, but parent mapping needs careful domain review.

DESIGN #5 — Security Boundary Corpus Runner
  Built from:     CMPSBL BLD-ENG-248 "Adversarial Input Generator" (ADVERSARIAL_CATALOG, generateAdversarialInputs)
  Plus host:      packages/cli/src/middlewares/__tests__/body-parser.test.ts; packages/@n8n/backend-network/src/ssrf/__tests__/ssrf-protection.service.test.ts; rate-limit boundary tests
  What it is:     A deterministic reusable corpus runner that applies the same security-shaped strings/payloads to multiple request boundaries and records which class caused each rejection/acceptance.
  New because:    n8n already property-tests selected algorithms; this makes a shared security-specific corpus a first-class fixture rather than one-off cases.
  Unlocks:        Cheap regression protection for encoding, Unicode, injection and overflow edge cases.
  Evidence path:  generateAdversarialInputs -> host adapter -> existing boundary -> assertion/result label -> CI
  Effort:         ~1–2 engineer-days integration (basis: 33 entry LOC + host adapters/fixtures/reporting; acquisition-equivalent size is hours under the kit band).
  Constraints:    BLD-ENG-248 Verified: C:ok I:ok B:— T:— S:scan; mode: direct; ports: none stated; constraints: none. B:— = Not claimed; T:— = Not claimed. A finite catalog is not a completeness claim or penetration test.
  Confidence:     high as a test-only quick win.

Scoreboard

The numbers, each with its basis

Step 1

What this repository already has

Host inventory was completed before catalog exposure. ZIP: 33,730 members = 27,048 regular files + 22 symlink entries + 6,660 directories. 19,556 TypeScript files; ~3,271,010 nonblank TS LOC.

Step 2

Five-axis screen of the candidate set

44 host-matched or false-positive-control rows were read in full after the host freeze. Twins are visible but never credited twice.

Step 3

Constraints this host imposes

Every retained row says constraints: none verbatim; host constraints still apply.

• Pre-purchase implementation source is unavailable; catalog metadata is the claim ceiling.
• Every retained row has T:— = Not claimed. S:scan is author self-scan, not independent audit.
• I:ok is empty-host installation/typecheck, not proof against n8n Node 24 / TypeScript 6 / Turbo build.
• Embedding in n8n is permitted after purchase under the product grant; preserve attribution and do not republish the CMPSBL delivery itself.
• Catalog rate limiters expose no shared-state/store port; do not claim a multi-main fix.
• Preserve n8n enterprise and permissive subpackage licence boundaries.

Step 4

Ranked recommendations

Findings, in fixed shape

[HARDEN] CMPSBL SUBP-06 "BULKHEAD" -> host: packages/cli/src/utils/ai-service-retry.ts (96 nonblank LOC); packages/@n8n/backend-network/src/http/undici/transport.ts (296 nonblank LOC)
  Why: Adds per-dependency concurrency and queue isolation where a timed-out or saturated external service can otherwise keep consuming process capacity. Start only on n8n-owned control-plane dependencies, not arbitrary user workflow traffic.
  Uses: BulkheadOptions, BulkheadRunOptions, Bulkhead, createBulkhead, BulkheadSubject, bulkheadContract, BulkheadRejection, BulkheadPolicy, RejectionReason +1 more
  Saves: ~1–3 engineer-weeks (basis: 539 entry LOC + compartment selection, queue/reject policy, concurrency accounting, snapshots and n8n-specific dependency-key tests; kit 400–1,200 LOC band).
  Verified: C:ok I:ok B:load T:— S:scan — C:ok = Compiles (batch tsgo --strict --noEmit); I:ok = installed with install.mjs into an empty host and type-checked there with tsgo --noEmit --skipLibCheck; B:load = entry point imports cleanly and exposes exports; no zero-arity behavior was fabricated; T:— = Not claimed (test coverage); S:scan = author's 13-rule deterministic static unsafe-construct sweep reported clean; not an independent security audit
  Mode: direct
  Ports to implement: none stated in catalog metadata

[SKIP] CMPSBL SUBP-05 "TRIAGE" -> host: packages/@n8n/backend-network/src/http/retryability.ts; packages/cli/src/utils/circuit-breaker.ts
  Why: n8n already has retryability classification, circuit breaking, queueing and domain-specific recovery; a 496-LOC general recovery ladder would overlap several mature systems.
  Uses: triageContract, createTriage, BUILT_IN_RULES, DEFAULT_CLASSIFICATION, classifyWith, normaliseKind, normaliseMessage, LadderRule, TriageOptions +8 more
  Saves: 0 engineer-days credited (basis: rejected candidate; no adoption value counted).
  Verified: C:ok I:ok B:load T:— S:scan — C:ok = Compiles (batch tsgo --strict --noEmit); I:ok = installed with install.mjs into an empty host and type-checked there with tsgo --noEmit --skipLibCheck; B:load = entry point imports cleanly and exposes exports; no zero-arity behavior was fabricated; T:— = Not claimed (test coverage); S:scan = author's 13-rule deterministic static unsafe-construct sweep reported clean; not an independent security audit
  Mode: direct
  Ports to implement: none stated in catalog metadata

[SKIP] CMPSBL SUBP-04 "HARNESS" -> host: packages/testing/**; packages/@n8n/scheduler/src/** tests
  Why: The repository already has very large unit/integration/property/E2E harnesses; a generic port-contract rig is useful only if CMPSBL ports are adopted later, not as host value by itself.
  Uses: httpContract, queueContract, secretsContract, stateContract, telemetryContract, defineContract, runContract, runSuite, PORT_CONTRACTS +9 more
  Saves: 0 engineer-days credited (basis: rejected candidate; no adoption value counted).
  Verified: C:ok I:ok B:load T:— S:scan — C:ok = Compiles (batch tsgo --strict --noEmit); I:ok = installed with install.mjs into an empty host and type-checked there with tsgo --noEmit --skipLibCheck; B:load = entry point imports cleanly and exposes exports; no zero-arity behavior was fabricated; T:— = Not claimed (test coverage); S:scan = author's 13-rule deterministic static unsafe-construct sweep reported clean; not an independent security audit
  Mode: direct
  Ports to implement: none stated in catalog metadata

[HARDEN] CMPSBL BLD-ENG-303 "retry-with-backoff" -> host: packages/core/src/execution-engine/workflow-execute.ts (2,590 nonblank LOC); packages/cli/src/utils/ai-service-retry.ts (96); packages/cli/src/modules/external-secrets.ee/retry-manager.service.ts (119)
  Why: Provides one small jittered, cancellable retry primitive that can replace selected duplicated helper logic. It must not silently replace user-visible workflow retry semantics, which are a product contract.
  Uses: RetryOptions, retry
  Saves: ~3–8 engineer-hours (basis: 55 entry LOC + backoff/jitter/predicate/AbortSignal concerns; kit <100 LOC band).
  Verified: C:ok I:ok B:— T:— S:scan — C:ok = Compiles (batch tsgo --strict --noEmit); I:ok = installed with install.mjs into an empty host and type-checked there with tsgo --noEmit --skipLibCheck; B:— = Not claimed (behavior); T:— = Not claimed (test coverage); S:scan = author's 13-rule deterministic static unsafe-construct sweep reported clean; not an independent security audit
  Mode: direct
  Ports to implement: none stated in catalog metadata

[HARDEN] CMPSBL BLD-ENG-266 "Retry Budget" -> host: packages/cli/src/utils/ai-service-retry.ts (96 nonblank LOC); packages/cli/src/modules/external-secrets.ee/retry-manager.service.ts (119 nonblank LOC)
  Why: Adds a refilling retry budget so separate retry loops share a pressure ceiling instead of each independently deciding that another retry is affordable. Particularly relevant to the external-secrets manager, whose scheduled retry loop has no attempt ceiling in the observed source.
  Uses: tryRetry, remainingBudget, configureBudget, getBudgetStats, resetBudget
  Saves: ~3–8 engineer-hours (basis: 56 entry LOC + budget/refill/observability concerns; kit <100 LOC band).
  Verified: C:ok I:ok B:— T:— S:scan — C:ok = Compiles (batch tsgo --strict --noEmit); I:ok = installed with install.mjs into an empty host and type-checked there with tsgo --noEmit --skipLibCheck; B:— = Not claimed (behavior); T:— = Not claimed (test coverage); S:scan = author's 13-rule deterministic static unsafe-construct sweep reported clean; not an independent security audit
  Mode: direct
  Ports to implement: none stated in catalog metadata

[HARDEN] CMPSBL BLD-ENG-009 "Tamper-Evident Chain" -> host: packages/cli/src/eventbus/message-event-bus-writer/message-event-bus-log-writer-worker.ts (130 nonblank LOC); packages/cli/src/eventbus/event-message-classes/abstract-event-message.ts (123)
  Why: n8n already emits rich execution/audit events, but the observed log writer appends ordinary JSON lines and no hash/Merkle linkage was found. A sidecar chain can make selected evidence tamper-evident without replacing the event bus.
  Uses: createAuditChain
  Saves: ~4–8 engineer-hours (basis: 67 entry LOC + canonicalization/hash-link/persistence/verification concerns; kit <100 LOC band).
  Verified: C:ok I:ok B:exec T:— S:scan — C:ok = Compiles (batch tsgo --strict --noEmit); I:ok = installed with install.mjs into an empty host and type-checked there with tsgo --noEmit --skipLibCheck; B:exec = entry point loads; zero-arity exports were executed twice and compared structurally; T:— = Not claimed (test coverage); S:scan = author's 13-rule deterministic static unsafe-construct sweep reported clean; not an independent security audit
  Mode: direct
  Ports to implement: none stated in catalog metadata

[SKIP] CMPSBL STIER-S-71 "Tamper-Evident Chain" -> host: eventbus evidence seam
  Why: Twin of the smaller carried BLD-ENG-009; counting both would double-count the same capability.
  Uses: createAuditChain
  Saves: 0 engineer-days credited (basis: rejected candidate; no adoption value counted).
  Verified: C:ok I:ok B:exec T:— S:scan — C:ok = Compiles (batch tsgo --strict --noEmit); I:ok = installed with install.mjs into an empty host and type-checked there with tsgo --noEmit --skipLibCheck; B:exec = entry point loads; zero-arity exports were executed twice and compared structurally; T:— = Not claimed (test coverage); S:scan = author's 13-rule deterministic static unsafe-construct sweep reported clean; not an independent security audit
  Mode: direct
  Ports to implement: none stated in catalog metadata

[HARDEN] CMPSBL BLD-ENG-105 "Forensic Replay" -> host: packages/cli/src/eventbus/** (2,204 nonblank LOC aggregate); execution/audit event classes
  Why: Turns captured audit/execution evidence into an incident reconstruction artifact. n8n has execution replay/test-run machinery, but no generic forensic event replay layer was found for the event-bus evidence stream.
  Uses: AuditEvent, ReplaySession, ReplayResult, ForensicReplayEngine
  Saves: ~1–3 engineer-days (basis: 103 entry LOC + event normalization/session slicing/replay/result concerns; kit 100–400 LOC band).
  Verified: C:ok I:ok B:— T:— S:scan — C:ok = Compiles (batch tsgo --strict --noEmit); I:ok = installed with install.mjs into an empty host and type-checked there with tsgo --noEmit --skipLibCheck; B:— = Not claimed (behavior); T:— = Not claimed (test coverage); S:scan = author's 13-rule deterministic static unsafe-construct sweep reported clean; not an independent security audit
  Mode: direct
  Ports to implement: none stated in catalog metadata

[HARDEN] CMPSBL STIER-S-81 "Compliance Attestation Generator" -> host: packages/cli/src/eventbus/**; workflow execution/audit records
  Why: Packages selected evidence into an attestation-style report with cryptographic evidence chaining. This complements n8n’s existing SBOM/attestation pipeline, which is supply-chain focused rather than per-workflow execution evidence.
  Uses: EvidenceItem, Attestation, generateAttestation, batchAttest
  Saves: ~3–8 engineer-hours (basis: 57 entry LOC + evidence selection/attestation batching/output contract concerns; kit <100 LOC band).
  Verified: C:ok I:ok B:load T:— S:scan — C:ok = Compiles (batch tsgo --strict --noEmit); I:ok = installed with install.mjs into an empty host and type-checked there with tsgo --noEmit --skipLibCheck; B:load = entry point imports cleanly and exposes exports; no zero-arity behavior was fabricated; T:— = Not claimed (test coverage); S:scan = author's 13-rule deterministic static unsafe-construct sweep reported clean; not an independent security audit
  Mode: direct
  Ports to implement: none stated in catalog metadata

[HARDEN] CMPSBL BLD-ENG-034 "Causal Event Propagation" -> host: packages/cli/src/eventbus/event-message-classes/abstract-event-message.ts (123 nonblank LOC); workflow/node/execution/queue/runner event families
  Why: The base event shape has id, timestamp, event name, message and payload, but no explicit parent/causal field was found. Causal propagation can add parent-child lineage across execution, node, queue and runner events without changing execution semantics.
  Uses: CausalEvent, PropagationRule, CausalChain, addRule, clearRules, propagate, getChain, getAllChains, getStats
  Saves: ~1–3 engineer-days (basis: 106 entry LOC + parent/rule/chain/stats concerns; kit 100–400 LOC band).
  Verified: C:ok I:ok B:— T:— S:scan — C:ok = Compiles (batch tsgo --strict --noEmit); I:ok = installed with install.mjs into an empty host and type-checked there with tsgo --noEmit --skipLibCheck; B:— = Not claimed (behavior); T:— = Not claimed (test coverage); S:scan = author's 13-rule deterministic static unsafe-construct sweep reported clean; not an independent security audit
  Mode: direct
  Ports to implement: none stated in catalog metadata

[SKIP] CMPSBL BLD-ENG-136 "Audit-Grade Decision Ledger" -> host: packages/cli/src/eventbus/**
  Why: A second decision ledger overlaps the carried tamper chain plus n8n’s existing event records; the report does not need both.
  Uses: LedgerEntry, AuditGradeDecisionLedger
  Saves: 0 engineer-days credited (basis: rejected candidate; no adoption value counted).
  Verified: C:ok I:ok B:load T:— S:scan — C:ok = Compiles (batch tsgo --strict --noEmit); I:ok = installed with install.mjs into an empty host and type-checked there with tsgo --noEmit --skipLibCheck; B:load = entry point imports cleanly and exposes exports; no zero-arity behavior was fabricated; T:— = Not claimed (test coverage); S:scan = author's 13-rule deterministic static unsafe-construct sweep reported clean; not an independent security audit
  Mode: direct
  Ports to implement: none stated in catalog metadata

[SKIP] CMPSBL STIER-S-CJ94 "Audit-Grade Decision Ledger" -> host: packages/cli/src/eventbus/**
  Why: Twin of BLD-ENG-136 and therefore excluded from sizing/value.
  Uses: LedgerEntry, AuditGradeDecisionLedger
  Saves: 0 engineer-days credited (basis: rejected candidate; no adoption value counted).
  Verified: C:ok I:ok B:load T:— S:scan — C:ok = Compiles (batch tsgo --strict --noEmit); I:ok = installed with install.mjs into an empty host and type-checked there with tsgo --noEmit --skipLibCheck; B:load = entry point imports cleanly and exposes exports; no zero-arity behavior was fabricated; T:— = Not claimed (test coverage); S:scan = author's 13-rule deterministic static unsafe-construct sweep reported clean; not an independent security audit
  Mode: direct
  Ports to implement: none stated in catalog metadata

[SKIP] CMPSBL BLD-ENG-075 "Self-Audit Loop" -> host: security/**; packages/cli/src/** audit services
  Why: n8n already has dedicated security/audit checks and operational telemetry; a generic self-audit registry was not tied to a missing host job strongly enough.
  Uses: AuditPolicy, AuditResult, AuditReport, registerPolicy, removePolicy, runAudit, getComplianceScore
  Saves: 0 engineer-days credited (basis: rejected candidate; no adoption value counted).
  Verified: C:ok I:ok B:exec T:— S:scan — C:ok = Compiles (batch tsgo --strict --noEmit); I:ok = installed with install.mjs into an empty host and type-checked there with tsgo --noEmit --skipLibCheck; B:exec = entry point loads; zero-arity exports were executed twice and compared structurally; T:— = Not claimed (test coverage); S:scan = author's 13-rule deterministic static unsafe-construct sweep reported clean; not an independent security audit
  Mode: direct
  Ports to implement: none stated in catalog metadata

[SKIP] CMPSBL STIER-S-78 "Self-Audit Loop" -> host: security/**
  Why: Twin of BLD-ENG-075; excluded with it.
  Uses: AuditPolicy, AuditResult, AuditReport, registerPolicy, removePolicy, runAudit, getComplianceScore
  Saves: 0 engineer-days credited (basis: rejected candidate; no adoption value counted).
  Verified: C:ok I:ok B:exec T:— S:scan — C:ok = Compiles (batch tsgo --strict --noEmit); I:ok = installed with install.mjs into an empty host and type-checked there with tsgo --noEmit --skipLibCheck; B:exec = entry point loads; zero-arity exports were executed twice and compared structurally; T:— = Not claimed (test coverage); S:scan = author's 13-rule deterministic static unsafe-construct sweep reported clean; not an independent security audit
  Mode: direct
  Ports to implement: none stated in catalog metadata

[SKIP] CMPSBL BLD-ENG-229 "Provenance Chain Verifier" -> host: security/sca/README.md
  Why: n8n already has a mature SBOM/VEX/attestation/signature pipeline for software supply-chain provenance. Event evidence is better served by the carried tamper chain.
  Uses: ProvenanceLink, ProvenanceVerdict, ProvenanceChainVerifier
  Saves: 0 engineer-days credited (basis: rejected candidate; no adoption value counted).
  Verified: C:ok I:ok B:— T:— S:scan — C:ok = Compiles (batch tsgo --strict --noEmit); I:ok = installed with install.mjs into an empty host and type-checked there with tsgo --noEmit --skipLibCheck; B:— = Not claimed (behavior); T:— = Not claimed (test coverage); S:scan = author's 13-rule deterministic static unsafe-construct sweep reported clean; not an independent security audit
  Mode: direct
  Ports to implement: none stated in catalog metadata

[SKIP] CMPSBL BLD-ENG-246 "Audit Log Ring" -> host: packages/cli/src/eventbus/**
  Why: A capped FIFO audit ring is weaker than the existing persisted event-bus writer and database-backed event/execution systems.
  Uses: AuditEventType, AuditEntry, AuditLogRing
  Saves: 0 engineer-days credited (basis: rejected candidate; no adoption value counted).
  Verified: C:ok I:ok B:— T:— S:scan — C:ok = Compiles (batch tsgo --strict --noEmit); I:ok = installed with install.mjs into an empty host and type-checked there with tsgo --noEmit --skipLibCheck; B:— = Not claimed (behavior); T:— = Not claimed (test coverage); S:scan = author's 13-rule deterministic static unsafe-construct sweep reported clean; not an independent security audit
  Mode: direct
  Ports to implement: none stated in catalog metadata

[HARDEN] CMPSBL BLD-ENG-248 "Adversarial Input Generator" -> host: packages/cli/src/middlewares/body-parser.ts (106 nonblank LOC); packages/@n8n/backend-network/src/ssrf/ssrf-protection.service.ts (309); packages/cli/src/services/rate-limit.service.ts (87) tests
  Why: Adds a reusable injection/overflow/encoding/Unicode corpus to existing property and integration tests. n8n already uses fast-check in several packages, so this is a security-boundary fixture pack, not a replacement testing strategy.
  Uses: ADVERSARIAL_CATALOG, generateAdversarialInputs
  Saves: ~2–6 engineer-hours (basis: 33 entry LOC + corpus classification/adapter/assertion concerns; kit <100 LOC band).
  Verified: C:ok I:ok B:— T:— S:scan — C:ok = Compiles (batch tsgo --strict --noEmit); I:ok = installed with install.mjs into an empty host and type-checked there with tsgo --noEmit --skipLibCheck; B:— = Not claimed (behavior); T:— = Not claimed (test coverage); S:scan = author's 13-rule deterministic static unsafe-construct sweep reported clean; not an independent security audit
  Mode: direct
  Ports to implement: none stated in catalog metadata

[SKIP] CMPSBL STIER-S-CJ132 "Adversarial Simulation Engine" -> host: packages/testing/**; packages/cli/src/evaluation.ee/**
  Why: The description is broad and behavior evidence is only B:load; n8n already has substantial security, integration and evaluation tests. The smaller adversarial input generator has a clearer attachment.
  Uses: AttackScenario, SimulationResult, AdversarialSimulation
  Saves: 0 engineer-days credited (basis: rejected candidate; no adoption value counted).
  Verified: C:ok I:ok B:load T:— S:scan — C:ok = Compiles (batch tsgo --strict --noEmit); I:ok = installed with install.mjs into an empty host and type-checked there with tsgo --noEmit --skipLibCheck; B:load = entry point imports cleanly and exposes exports; no zero-arity behavior was fabricated; T:— = Not claimed (test coverage); S:scan = author's 13-rule deterministic static unsafe-construct sweep reported clean; not an independent security audit
  Mode: direct
  Ports to implement: none stated in catalog metadata

[SKIP] CMPSBL BLD-ENG-050 "Shadow Run Environment" -> host: packages/cli/src/evaluation.ee/**
  Why: n8n already ships a substantial enterprise Evaluation/test-run system that snapshots workflow versions and executes test runs; a generic shadow-run environment would duplicate it.
  Uses: ShadowRun, ShadowAssertion, createShadowRun, startRun, addAssertion, addLog, completeRun, getRun, listRuns
  Saves: 0 engineer-days credited (basis: rejected candidate; no adoption value counted).
  Verified: C:ok I:ok B:load T:— S:scan — C:ok = Compiles (batch tsgo --strict --noEmit); I:ok = installed with install.mjs into an empty host and type-checked there with tsgo --noEmit --skipLibCheck; B:load = entry point imports cleanly and exposes exports; no zero-arity behavior was fabricated; T:— = Not claimed (test coverage); S:scan = author's 13-rule deterministic static unsafe-construct sweep reported clean; not an independent security audit
  Mode: direct
  Ports to implement: none stated in catalog metadata

[SKIP] CMPSBL STIER-S-124 "Shadow Run Environment" -> host: packages/cli/src/evaluation.ee/**
  Why: Twin of BLD-ENG-050 and rejected for the same duplication.
  Uses: ShadowRun, ShadowAssertion, createShadowRun, startRun, addAssertion, addLog, completeRun, getRun, listRuns
  Saves: 0 engineer-days credited (basis: rejected candidate; no adoption value counted).
  Verified: C:ok I:ok B:load T:— S:scan — C:ok = Compiles (batch tsgo --strict --noEmit); I:ok = installed with install.mjs into an empty host and type-checked there with tsgo --noEmit --skipLibCheck; B:load = entry point imports cleanly and exposes exports; no zero-arity behavior was fabricated; T:— = Not claimed (test coverage); S:scan = author's 13-rule deterministic static unsafe-construct sweep reported clean; not an independent security audit
  Mode: direct
  Ports to implement: none stated in catalog metadata

[SKIP] CMPSBL SUBP-07 "CEILING" -> host: packages/cli/src/modules/instance-ai/instance-ai-credit.service.ts
  Why: Instance AI already has authoritative credit accounting, quotas, quota locking, usage claims and UI/telemetry updates. A second generic ceiling is not a clean incremental feature there.
  Uses: CeilingOptions, Ceiling, createCeiling, CeilingSubject, ceilingContract, METERS, Meter, MeterMap, CeilingSpec +5 more
  Saves: 0 engineer-days credited (basis: rejected candidate; no adoption value counted).
  Verified: C:ok I:ok B:load T:— S:scan — C:ok = Compiles (batch tsgo --strict --noEmit); I:ok = installed with install.mjs into an empty host and type-checked there with tsgo --noEmit --skipLibCheck; B:load = entry point imports cleanly and exposes exports; no zero-arity behavior was fabricated; T:— = Not claimed (test coverage); S:scan = author's 13-rule deterministic static unsafe-construct sweep reported clean; not an independent security audit
  Mode: direct
  Ports to implement: none stated in catalog metadata

[SKIP] CMPSBL BLD-ENG-121 "Autonomy Budget Manager" -> host: packages/cli/src/modules/instance-ai/instance-ai-credit.service.ts
  Why: n8n already meters/claims Instance AI credits and enforces quota locks; adding a parallel autonomy budget manager risks conflicting sources of truth.
  Uses: AutonomyBudget, BudgetAlert, SpendRequest, SpendResult, AutonomyBudgetManager
  Saves: 0 engineer-days credited (basis: rejected candidate; no adoption value counted).
  Verified: C:ok I:ok B:load T:— S:scan — C:ok = Compiles (batch tsgo --strict --noEmit); I:ok = installed with install.mjs into an empty host and type-checked there with tsgo --noEmit --skipLibCheck; B:load = entry point imports cleanly and exposes exports; no zero-arity behavior was fabricated; T:— = Not claimed (test coverage); S:scan = author's 13-rule deterministic static unsafe-construct sweep reported clean; not an independent security audit
  Mode: direct
  Ports to implement: none stated in catalog metadata

[SKIP] CMPSBL STIER-S-CJ79 "Autonomy Budget Manager" -> host: Instance AI credit system
  Why: Twin of BLD-ENG-121 and rejected for the same reason.
  Uses: AutonomyBudget, BudgetAlert, SpendRequest, SpendResult, AutonomyBudgetManager
  Saves: 0 engineer-days credited (basis: rejected candidate; no adoption value counted).
  Verified: C:ok I:ok B:load T:— S:scan — C:ok = Compiles (batch tsgo --strict --noEmit); I:ok = installed with install.mjs into an empty host and type-checked there with tsgo --noEmit --skipLibCheck; B:load = entry point imports cleanly and exposes exports; no zero-arity behavior was fabricated; T:— = Not claimed (test coverage); S:scan = author's 13-rule deterministic static unsafe-construct sweep reported clean; not an independent security audit
  Mode: direct
  Ports to implement: none stated in catalog metadata

[SKIP] CMPSBL BLD-ENG-049 "Realtime Cost Attribution" -> host: packages/@n8n/ai-utilities/src/utils/n8n-llm-tracing.ts
  Why: n8n AI tracing already records token totals and can record cost totals; generic cost attribution is not sufficiently novel.
  Uses: CostEntry, ModuleCostSummary, recordCost, getModuleCosts, getTotalCost, getLedgerSize
  Saves: 0 engineer-days credited (basis: rejected candidate; no adoption value counted).
  Verified: C:ok I:ok B:— T:— S:scan — C:ok = Compiles (batch tsgo --strict --noEmit); I:ok = installed with install.mjs into an empty host and type-checked there with tsgo --noEmit --skipLibCheck; B:— = Not claimed (behavior); T:— = Not claimed (test coverage); S:scan = author's 13-rule deterministic static unsafe-construct sweep reported clean; not an independent security audit
  Mode: direct
  Ports to implement: none stated in catalog metadata

[SKIP] CMPSBL BLD-ENG-006 "Cost-Aware Routing Engine" -> host: AI/model nodes and Instance AI model resolution
  Why: Automatic cost-aware rerouting would violate workflows that explicitly select a model/provider/credential. This is a product-semantic mismatch for generic workflow execution.
  Uses: createCostRouter
  Saves: 0 engineer-days credited (basis: rejected candidate; no adoption value counted).
  Verified: C:ok I:ok B:load T:— S:scan — C:ok = Compiles (batch tsgo --strict --noEmit); I:ok = installed with install.mjs into an empty host and type-checked there with tsgo --noEmit --skipLibCheck; B:load = entry point imports cleanly and exposes exports; no zero-arity behavior was fabricated; T:— = Not claimed (test coverage); S:scan = author's 13-rule deterministic static unsafe-construct sweep reported clean; not an independent security audit
  Mode: direct
  Ports to implement: none stated in catalog metadata

[SKIP] CMPSBL BLD-ENG-047 "Circuit Breaker Fabric" -> host: packages/cli/src/utils/circuit-breaker.ts
  Why: n8n already has a mature circuit breaker with closed/open/half-open states, sliding failure window and half-open concurrency. A breaker fabric is not justified by the observed gap.
  Uses: BreakerState, CircuitBreaker, createBreaker, recordSuccess, recordFailure, canExecute, getBreaker, getAllBreakers, resetBreaker
  Saves: 0 engineer-days credited (basis: rejected candidate; no adoption value counted).
  Verified: C:ok I:ok B:exec T:— S:scan — C:ok = Compiles (batch tsgo --strict --noEmit); I:ok = installed with install.mjs into an empty host and type-checked there with tsgo --noEmit --skipLibCheck; B:exec = entry point loads; zero-arity exports were executed twice and compared structurally; T:— = Not claimed (test coverage); S:scan = author's 13-rule deterministic static unsafe-construct sweep reported clean; not an independent security audit
  Mode: direct
  Ports to implement: none stated in catalog metadata

[SKIP] CMPSBL BLD-ENG-011 "Circuit Breaker" -> host: packages/cli/src/utils/circuit-breaker.ts
  Why: Direct duplicate of capability n8n already owns in a stronger host-specific implementation.
  Uses: CircuitState, CircuitBreakerConfig, CircuitStats, createCircuitBreaker, createBreakerPanel
  Saves: 0 engineer-days credited (basis: rejected candidate; no adoption value counted).
  Verified: C:ok I:ok B:load T:— S:scan — C:ok = Compiles (batch tsgo --strict --noEmit); I:ok = installed with install.mjs into an empty host and type-checked there with tsgo --noEmit --skipLibCheck; B:load = entry point imports cleanly and exposes exports; no zero-arity behavior was fabricated; T:— = Not claimed (test coverage); S:scan = author's 13-rule deterministic static unsafe-construct sweep reported clean; not an independent security audit
  Mode: direct
  Ports to implement: none stated in catalog metadata

[SKIP] CMPSBL BLD-ENG-045 "Provider Health Monitor" -> host: existing telemetry/circuit-breaker surfaces
  Why: Provider health is already represented through host telemetry/circuit state in the relevant control planes; no generic provider selector should override user-selected integrations.
  Uses: ProviderSample, ProviderHealth, recordSample, getProviderHealth, getAllProviderHealth, getBestProvider
  Saves: 0 engineer-days credited (basis: rejected candidate; no adoption value counted).
  Verified: C:ok I:ok B:exec T:— S:scan — C:ok = Compiles (batch tsgo --strict --noEmit); I:ok = installed with install.mjs into an empty host and type-checked there with tsgo --noEmit --skipLibCheck; B:exec = entry point loads; zero-arity exports were executed twice and compared structurally; T:— = Not claimed (test coverage); S:scan = author's 13-rule deterministic static unsafe-construct sweep reported clean; not an independent security audit
  Mode: direct
  Ports to implement: none stated in catalog metadata

[SKIP] CMPSBL BLD-ENG-078 "Fallback Chain Architect" -> host: workflow node/provider semantics
  Why: Automatic fallback chains are unsafe for arbitrary workflow integrations because providers are not semantically interchangeable and users explicitly choose credentials/providers.
  Uses: ProviderSpec, FallbackChain, buildFallbackChain
  Saves: 0 engineer-days credited (basis: rejected candidate; no adoption value counted).
  Verified: C:ok I:ok B:load T:— S:scan — C:ok = Compiles (batch tsgo --strict --noEmit); I:ok = installed with install.mjs into an empty host and type-checked there with tsgo --noEmit --skipLibCheck; B:load = entry point imports cleanly and exposes exports; no zero-arity behavior was fabricated; T:— = Not claimed (test coverage); S:scan = author's 13-rule deterministic static unsafe-construct sweep reported clean; not an independent security audit
  Mode: direct
  Ports to implement: none stated in catalog metadata

[SKIP] CMPSBL STIER-S-101 "Connector Orchestration Engine" -> host: packages/nodes-base/**; packages/core/**
  Why: n8n itself is already a connector orchestration product with node lifecycle, credentials, retries and 1,500+ integrations; importing a generic connector orchestrator would mostly duplicate the host.
  Uses: Connector, registerConnector, pingConnector, getConnector, listConnectors, removeConnector, executeWithRetry
  Saves: 0 engineer-days credited (basis: rejected candidate; no adoption value counted).
  Verified: C:ok I:ok B:load T:— S:scan — C:ok = Compiles (batch tsgo --strict --noEmit); I:ok = installed with install.mjs into an empty host and type-checked there with tsgo --noEmit --skipLibCheck; B:load = entry point imports cleanly and exposes exports; no zero-arity behavior was fabricated; T:— = Not claimed (test coverage); S:scan = author's 13-rule deterministic static unsafe-construct sweep reported clean; not an independent security audit
  Mode: direct
  Ports to implement: none stated in catalog metadata

[SKIP] CMPSBL BLD-ENG-109 "External API Rate Limiter" -> host: packages/cli/src/services/rate-limit.service.ts; multi-main deployment
  Why: The catalog API exposes no shared-state/store port, so pre-purchase metadata cannot establish process-global/provider-global quota enforcement. It does not solve the observed multi-main state seam.
  Uses: RateLimitConfig, RateLimitResult, ExternalAPIRateLimiter
  Saves: 0 engineer-days credited (basis: rejected candidate; no adoption value counted).
  Verified: C:ok I:ok B:load T:— S:scan — C:ok = Compiles (batch tsgo --strict --noEmit); I:ok = installed with install.mjs into an empty host and type-checked there with tsgo --noEmit --skipLibCheck; B:load = entry point imports cleanly and exposes exports; no zero-arity behavior was fabricated; T:— = Not claimed (test coverage); S:scan = author's 13-rule deterministic static unsafe-construct sweep reported clean; not an independent security audit
  Mode: direct
  Ports to implement: none stated in catalog metadata

[SKIP] CMPSBL STIER-S-133 "External API Rate Limiter" -> host: multi-main rate-limit seam
  Why: Twin of BLD-ENG-109 and has the same missing distributed-state evidence; excluded.
  Uses: RateLimitConfig, RateLimitResult, ExternalAPIRateLimiter
  Saves: 0 engineer-days credited (basis: rejected candidate; no adoption value counted).
  Verified: C:ok I:ok B:load T:— S:scan — C:ok = Compiles (batch tsgo --strict --noEmit); I:ok = installed with install.mjs into an empty host and type-checked there with tsgo --noEmit --skipLibCheck; B:load = entry point imports cleanly and exposes exports; no zero-arity behavior was fabricated; T:— = Not claimed (test coverage); S:scan = author's 13-rule deterministic static unsafe-construct sweep reported clean; not an independent security audit
  Mode: direct
  Ports to implement: none stated in catalog metadata

[SKIP] CMPSBL BLD-ENG-012 "Rate Limiter" -> host: packages/cli/src/services/rate-limit.service.ts
  Why: No shared-state/store port is exposed in the catalog API; it cannot honestly replace a multi-process limiter where global counters matter.
  Uses: Strategy, RateLimitConfig, RateLimitResult, createRateLimiter
  Saves: 0 engineer-days credited (basis: rejected candidate; no adoption value counted).
  Verified: C:ok I:ok B:load T:— S:scan — C:ok = Compiles (batch tsgo --strict --noEmit); I:ok = installed with install.mjs into an empty host and type-checked there with tsgo --noEmit --skipLibCheck; B:load = entry point imports cleanly and exposes exports; no zero-arity behavior was fabricated; T:— = Not claimed (test coverage); S:scan = author's 13-rule deterministic static unsafe-construct sweep reported clean; not an independent security audit
  Mode: direct
  Ports to implement: none stated in catalog metadata

[SKIP] CMPSBL STIER-EXTRA-03 "Rate Limiter" -> host: packages/cli/src/services/rate-limit.service.ts
  Why: Twin of BLD-ENG-012; multi-strategy choice does not solve the distributed-state problem.
  Uses: Strategy, RateLimitConfig, RateLimitResult, createRateLimiter
  Saves: 0 engineer-days credited (basis: rejected candidate; no adoption value counted).
  Verified: C:ok I:ok B:load T:— S:scan — C:ok = Compiles (batch tsgo --strict --noEmit); I:ok = installed with install.mjs into an empty host and type-checked there with tsgo --noEmit --skipLibCheck; B:load = entry point imports cleanly and exposes exports; no zero-arity behavior was fabricated; T:— = Not claimed (test coverage); S:scan = author's 13-rule deterministic static unsafe-construct sweep reported clean; not an independent security audit
  Mode: direct
  Ports to implement: none stated in catalog metadata

[SKIP] CMPSBL STIER-S-119 "Quota Intelligence Engine" -> host: third-party integration quota surfaces
  Why: Quota prediction requires authoritative quota limits and shared consumption state per provider. The catalog row does not expose the host adapters needed to establish that globally.
  Uses: QuotaConfig, QuotaWarning, registerQuota, consumeQuota, getWarning, listQuotas
  Saves: 0 engineer-days credited (basis: rejected candidate; no adoption value counted).
  Verified: C:ok I:ok B:exec T:— S:scan — C:ok = Compiles (batch tsgo --strict --noEmit); I:ok = installed with install.mjs into an empty host and type-checked there with tsgo --noEmit --skipLibCheck; B:exec = entry point loads; zero-arity exports were executed twice and compared structurally; T:— = Not claimed (test coverage); S:scan = author's 13-rule deterministic static unsafe-construct sweep reported clean; not an independent security audit
  Mode: direct
  Ports to implement: none stated in catalog metadata

[SKIP] CMPSBL BLD-ENG-089 "Quota Intelligence" -> host: third-party integration quota surfaces
  Why: Smaller twin/variant of quota intelligence; same authoritative-state problem.
  Uses: QuotaConfig, QuotaWarning, registerQuota, consumeQuota, getWarning, listQuotas
  Saves: 0 engineer-days credited (basis: rejected candidate; no adoption value counted).
  Verified: C:ok I:ok B:— T:— S:scan — C:ok = Compiles (batch tsgo --strict --noEmit); I:ok = installed with install.mjs into an empty host and type-checked there with tsgo --noEmit --skipLibCheck; B:— = Not claimed (behavior); T:— = Not claimed (test coverage); S:scan = author's 13-rule deterministic static unsafe-construct sweep reported clean; not an independent security audit
  Mode: direct
  Ports to implement: none stated in catalog metadata

[SKIP] CMPSBL BLD-ENG-215 "Contract Evolution Mediator" -> host: nodes-base node versions; workflow migrations
  Why: n8n already has node typeVersion/versioned descriptions and migration machinery. A generic contract mediator was not shown to improve that host-specific versioning model.
  Uses: ContractVersion, Contract, ContractEvolutionMediator
  Saves: 0 engineer-days credited (basis: rejected candidate; no adoption value counted).
  Verified: C:ok I:ok B:load T:— S:scan — C:ok = Compiles (batch tsgo --strict --noEmit); I:ok = installed with install.mjs into an empty host and type-checked there with tsgo --noEmit --skipLibCheck; B:load = entry point imports cleanly and exposes exports; no zero-arity behavior was fabricated; T:— = Not claimed (test coverage); S:scan = author's 13-rule deterministic static unsafe-construct sweep reported clean; not an independent security audit
  Mode: direct
  Ports to implement: none stated in catalog metadata

[SKIP] CMPSBL BLD-ENG-202 "Event-Driven Integration Mesh" -> host: packages/cli/src/eventbus/**; scaling/queue system
  Why: n8n already has event-bus persistence plus Redis/Bull queueing/scaling. A second generic pub-sub mesh would create parallel delivery semantics.
  Uses: EventStream, DeadLetter, EventDrivenIntegrationMesh
  Saves: 0 engineer-days credited (basis: rejected candidate; no adoption value counted).
  Verified: C:ok I:ok B:load T:— S:scan — C:ok = Compiles (batch tsgo --strict --noEmit); I:ok = installed with install.mjs into an empty host and type-checked there with tsgo --noEmit --skipLibCheck; B:load = entry point imports cleanly and exposes exports; no zero-arity behavior was fabricated; T:— = Not claimed (test coverage); S:scan = author's 13-rule deterministic static unsafe-construct sweep reported clean; not an independent security audit
  Mode: direct
  Ports to implement: none stated in catalog metadata

[SKIP] CMPSBL BLD-ENG-096 "Webhook Reliability" -> host: packages/cli/src/webhooks/**; Webhook node
  Why: The row describes webhook delivery reliability, while n8n’s primary generic Webhook node is an inbound trigger. Applying outbound-delivery semantics there would be a category error.
  Uses: WebhookEvent, DeliveryAttempt, WebhookConfig, WebhookReliabilityEngine
  Saves: 0 engineer-days credited (basis: rejected candidate; no adoption value counted).
  Verified: C:ok I:ok B:— T:— S:scan — C:ok = Compiles (batch tsgo --strict --noEmit); I:ok = installed with install.mjs into an empty host and type-checked there with tsgo --noEmit --skipLibCheck; B:— = Not claimed (behavior); T:— = Not claimed (test coverage); S:scan = author's 13-rule deterministic static unsafe-construct sweep reported clean; not an independent security audit
  Mode: direct
  Ports to implement: none stated in catalog metadata

[SKIP] CMPSBL STIER-S-114 "Webhook Reliability Engine" -> host: packages/cli/src/webhooks/**
  Why: Richer twin/variant of outbound webhook delivery; still the wrong trust/direction model for n8n’s inbound generic webhook surface.
  Uses: WebhookEvent, DeliveryAttempt, WebhookConfig, WebhookStore, DeliveryDisposition, WebhookReliabilityEngine
  Saves: 0 engineer-days credited (basis: rejected candidate; no adoption value counted).
  Verified: C:ok I:ok B:load T:— S:scan — C:ok = Compiles (batch tsgo --strict --noEmit); I:ok = installed with install.mjs into an empty host and type-checked there with tsgo --noEmit --skipLibCheck; B:load = entry point imports cleanly and exposes exports; no zero-arity behavior was fabricated; T:— = Not claimed (test coverage); S:scan = author's 13-rule deterministic static unsafe-construct sweep reported clean; not an independent security audit
  Mode: direct
  Ports to implement: none stated in catalog metadata

[SKIP] CMPSBL BLD-ENG-264 "Idempotency Keys" -> host: packages/core/src/data-deduplication-service.ts
  Why: Its catalog behavior is an in-memory TTL/in-flight cache, while n8n already has persisted deduplication for selected workloads. It is insufficient as a cross-process generic webhook idempotency guarantee.
  Uses: withIdempotency, hasIdempotencyKey, clearIdempotencyKey, getIdempotencyStats
  Saves: 0 engineer-days credited (basis: rejected candidate; no adoption value counted).
  Verified: C:ok I:ok B:— T:— S:scan — C:ok = Compiles (batch tsgo --strict --noEmit); I:ok = installed with install.mjs into an empty host and type-checked there with tsgo --noEmit --skipLibCheck; B:— = Not claimed (behavior); T:— = Not claimed (test coverage); S:scan = author's 13-rule deterministic static unsafe-construct sweep reported clean; not an independent security audit
  Mode: direct
  Ports to implement: none stated in catalog metadata

[SKIP] CMPSBL STIER-S-ACC02 "Adaptive Rate Limiting Engine" -> host: packages/cli/src/services/rate-limit.service.ts
  Why: Adaptive local limiting does not address the observed multi-main/shared-counter seam; no shared-state port appears in the catalog API.
  Uses: RateLimiter, createLimiter, tryConsume, adaptLimit, getLimiter, listLimiters
  Saves: 0 engineer-days credited (basis: rejected candidate; no adoption value counted).
  Verified: C:ok I:ok B:exec T:— S:scan — C:ok = Compiles (batch tsgo --strict --noEmit); I:ok = installed with install.mjs into an empty host and type-checked there with tsgo --noEmit --skipLibCheck; B:exec = entry point loads; zero-arity exports were executed twice and compared structurally; T:— = Not claimed (test coverage); S:scan = author's 13-rule deterministic static unsafe-construct sweep reported clean; not an independent security audit
  Mode: direct
  Ports to implement: none stated in catalog metadata

[SKIP] CMPSBL BLD-ENG-066 "Adaptive Rate Limiting" -> host: packages/cli/src/services/rate-limit.service.ts
  Why: Same issue as the adaptive S-tier limiter: useful local policy, but no evidence it can coordinate counters across main processes.
  Uses: RateLimiter, createLimiter, tryConsume, adaptLimit, getLimiter, listLimiters
  Saves: 0 engineer-days credited (basis: rejected candidate; no adoption value counted).
  Verified: C:ok I:ok B:— T:— S:scan — C:ok = Compiles (batch tsgo --strict --noEmit); I:ok = installed with install.mjs into an empty host and type-checked there with tsgo --noEmit --skipLibCheck; B:— = Not claimed (behavior); T:— = Not claimed (test coverage); S:scan = author's 13-rule deterministic static unsafe-construct sweep reported clean; not an independent security audit
  Mode: direct
  Ports to implement: none stated in catalog metadata

[SKIP] CMPSBL SUBSYS-04 "Shadow Mesh" -> host: packages/cli/src/evaluation.ee/**
  Why: n8n already has workflow Evaluation/test-run machinery and partial/manual execution; a 14-LOC traffic-mirroring subsystem row is not enough pre-purchase evidence to displace that system.
  Uses: generateAdversarialInputs, runShadowProbe, runAllShadowProbes, runShadowBatch, startShadowScheduler, stopShadowScheduler, getShadowMeshAnalytics, runShadowBuild, runAllShadowBuilds +6 more
  Saves: 0 engineer-days credited (basis: rejected candidate; no adoption value counted).
  Verified: C:proj(14) I:ok B:— T:— S:scan — C:proj(14) = compiles under project settings but strict isolated pass records the stated error count; not C:ok; I:ok = installed with install.mjs into an empty host and type-checked there with tsgo --noEmit --skipLibCheck; B:— = Not claimed (behavior); T:— = Not claimed (test coverage); S:scan = author's 13-rule deterministic static unsafe-construct sweep reported clean; not an independent security audit
  Mode: direct
  Ports to implement: none stated in catalog metadata

Gap map

Do not touch

• existing circuit breaker
• SSRF defenses
• public API-key lifecycle
• credential encryption/key manager except host-specific TODOs
• security/sca SBOM/VEX pipeline
• enterprise/AI workflow evaluation systems
• Instance AI credit system
• RedisLockService
• user-facing workflow retryOnFail semantics

Quick wins

• Event-chain sidecar
• Retry Budget Adapter on external-secrets/internal retry pressure
• Security Boundary Corpus Runner in tests

Step 6

Build versus buy

Assumptions

• US$150/hour senior engineer, 40h/week for conversion only.
• 1,016 total adopted entry LOC; closure LOC is not used as capability size.
• ~30 correctness concerns across the eight rows.
• TypeScript direct still needs source/security review, adapters, tests, flags and licence review.
• No funded roadmap evidence; Bucket A remains zero.

What the team would never have arrived at alone

Not claimed. n8n already demonstrates sophisticated independent engineering. The defensible value is prebuilt implementation and synthesis, not conceptual exclusivity.

Is this a good trade at the stated price?

Yes, technically, if either major capability family is wanted. Even low capability value minus high integration and the US$499 fee remains positive by about US$2,501. If post-purchase source review weakens the retained claims or n8n does not want these capabilities, the recommendation becomes HOLD regardless of the low licence price.

Step 7

90-day roadmap, assuming it was bought

Do not start: breaker/SSRF/API-key/encryption/evaluation/credit replacements; local catalog limiters as a multi-main fix; automatic provider fallback for user workflows.

Discovery Vault — provisional, kept out of every total

Machine-synthesised discoveries worth reading after purchase

All 6,479 Discovery Vault rows were screened by index metadata. No Discovery source/API/LOC/verification is credited pre-purchase.

Step 8 — appended, never merged into the body

Self-correction addendum

The final inventory check confirms that the evaluation modified no original n8n source. The two material corrections were in the evaluator/kit layer rather than the host: the archive member count needed symlink/directory qualification, and the first draft incorrectly created a public-source written-permission gate that the authoritative CMPSBL licence does not impose on a substantial product such as n8n.

Kit edge-case result: the current licence files and AGENTS.md must override stale Phase-1/template/example licensing prose, and layer-class attachment rules must override catalog cells that say Mode: direct. These inconsistencies did not change the eight retained n8n components, but they can change a future evaluator's verdict if read literally.

Claim ceiling

What is and is not verified

Extracted from systems the author has run in production since August 2025. Operationally exercised, not independently security-audited.

In pre-purchase evaluation, verification axes quoted in this report come verbatim from the generated AGENT-HANDOFF-CATALOG.md, which is the pre-purchase source of truth; in a full delivery, artifacts/<ID>/artifact-manifest.json is authoritative and overrides it. Every retained component records T:—, meaning Not claimed test coverage, and no component is claimed here to have had an independent third-party security review. Static scan results are the author's own.

• BLD-ENG-009 Tamper-Evident Chain — C:ok I:ok B:exec T:— S:scan — C:ok = Compiles (batch tsgo --strict --noEmit); I:ok = installed with install.mjs into an empty host and type-checked there with tsgo --noEmit --skipLibCheck; B:exec = entry point loads; zero-arity exports were executed twice and compared structurally; T:— = Not claimed (test coverage); S:scan = author's 13-rule deterministic static unsafe-construct sweep reported clean; not an independent security audit
• SUBP-06 BULKHEAD — C:ok I:ok B:load T:— S:scan — C:ok = Compiles (batch tsgo --strict --noEmit); I:ok = installed with install.mjs into an empty host and type-checked there with tsgo --noEmit --skipLibCheck; B:load = entry point imports cleanly and exposes exports; no zero-arity behavior was fabricated; T:— = Not claimed (test coverage); S:scan = author's 13-rule deterministic static unsafe-construct sweep reported clean; not an independent security audit
• BLD-ENG-266 Retry Budget — C:ok I:ok B:— T:— S:scan — C:ok = Compiles (batch tsgo --strict --noEmit); I:ok = installed with install.mjs into an empty host and type-checked there with tsgo --noEmit --skipLibCheck; B:— = Not claimed (behavior); T:— = Not claimed (test coverage); S:scan = author's 13-rule deterministic static unsafe-construct sweep reported clean; not an independent security audit
• BLD-ENG-105 Forensic Replay — C:ok I:ok B:— T:— S:scan — C:ok = Compiles (batch tsgo --strict --noEmit); I:ok = installed with install.mjs into an empty host and type-checked there with tsgo --noEmit --skipLibCheck; B:— = Not claimed (behavior); T:— = Not claimed (test coverage); S:scan = author's 13-rule deterministic static unsafe-construct sweep reported clean; not an independent security audit
• STIER-S-81 Compliance Attestation Generator — C:ok I:ok B:load T:— S:scan — C:ok = Compiles (batch tsgo --strict --noEmit); I:ok = installed with install.mjs into an empty host and type-checked there with tsgo --noEmit --skipLibCheck; B:load = entry point imports cleanly and exposes exports; no zero-arity behavior was fabricated; T:— = Not claimed (test coverage); S:scan = author's 13-rule deterministic static unsafe-construct sweep reported clean; not an independent security audit
• BLD-ENG-034 Causal Event Propagation — C:ok I:ok B:— T:— S:scan — C:ok = Compiles (batch tsgo --strict --noEmit); I:ok = installed with install.mjs into an empty host and type-checked there with tsgo --noEmit --skipLibCheck; B:— = Not claimed (behavior); T:— = Not claimed (test coverage); S:scan = author's 13-rule deterministic static unsafe-construct sweep reported clean; not an independent security audit
• BLD-ENG-303 retry-with-backoff — C:ok I:ok B:— T:— S:scan — C:ok = Compiles (batch tsgo --strict --noEmit); I:ok = installed with install.mjs into an empty host and type-checked there with tsgo --noEmit --skipLibCheck; B:— = Not claimed (behavior); T:— = Not claimed (test coverage); S:scan = author's 13-rule deterministic static unsafe-construct sweep reported clean; not an independent security audit
• BLD-ENG-248 Adversarial Input Generator — C:ok I:ok B:— T:— S:scan — C:ok = Compiles (batch tsgo --strict --noEmit); I:ok = installed with install.mjs into an empty host and type-checked there with tsgo --noEmit --skipLibCheck; B:— = Not claimed (behavior); T:— = Not claimed (test coverage); S:scan = author's 13-rule deterministic static unsafe-construct sweep reported clean; not an independent security audit

Governed runtime status: Not claimed. No CMPSBL implementation was installed or executed in n8n during this pre-purchase evaluation.