fix: add observability consistently across NFR stages

[jeromevdl]

Summary

Fixes #398 — NFR categories were inconsistent across stages: nfr-requirements.md listed 5 categories (Performance, Security, Scalability, Reliability, Observability) while requirements-analysis.md, nfr-design.md, and infrastructure-design.md only listed 4 (missing Observability).

Changes

Core stage files (core/aidlc-common/stages/)

• inception/requirements-analysis.md: Added "observability" to the NFR category list in Step 6 (completeness analysis) and Step 10 (generate requirements)
• construction/nfr-requirements.md: Added observability-requirements artifact to produces:, outputs:, and Step 6; updated condition: to list all 5 NFR categories
• construction/nfr-design.md: Added observability-design artifact to produces:, outputs:, Step 5, and Step 6; added observability-requirements to consumes:; updated upstream-coverage sensor list
• construction/infrastructure-design.md: Added observability-design to consumes: list and upstream-coverage sensor description

Generated dist/

• Regenerated all 4 harness trees via bun scripts/package.ts (pass --check)

Tests

• Regenerated tests/fixtures/designer-export/export.json golden fixture
• Updated tests/integration/t66.test.ts element count assertion (122→124 artifacts, 11→13 agents) and test label

Artifact flow (after fix)

requirements-analysis (observability in NFR dimensions)
    → nfr-requirements (produces: observability-requirements)
        → nfr-design (consumes: observability-requirements, produces: observability-design)
            → infrastructure-design (consumes: observability-design, produces: monitoring-design)

---

By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of the project license.

[apackeer]

Thanks for the fix; the shape is right and the artifact graph is internally consistent. Every new produces has a consumer, and every new required consumes has a producer in every scope that runs the consumer stage. Requesting changes for rebase and same-commit hygiene the current v2 requires.

Blockers

1. Rebase required. Base is 2.1.3 (c95070e); v2 is now 2.2.0 (b67798c) with six intervening commits (2.1.4 - 2.2.0). git merge-tree origin/v2 refs/pull/404/head reports a hard conflict on tests/integration/t66.test.ts; the stage .md files and dist trees auto-merge textually but every dist/*/tools/data/stage-graph.json and tests/fixtures/designer-export/export.json needs a real regeneration on the rebased tree.

2. Version trio missing. core/tools/aidlc-version.ts still reads 2.1.3, and there is no matching heading in CHANGELOG.md or update to the README badge. AGENTS.md Changelog Policy requires all three in the same commit and tests/unit/t68-version-changelog-sync.test.ts enforces it. Versions 2.2.1 - 2.2.13 are already claimed by open PRs, so 2.2.14 is the next free slot (re-bump if a sibling merges first).

3. t66 agent count stale. The PR sets artifacts=123, agents=13 in tests/integration/t66.test.ts; v2 has artifacts=122, agents=14 after the composer agent landed in 2.2.0. On rebase the target is stages=32, scopes=9, artifacts=123, agents=14, with both expect(...).toBe(...) lines matching. Your artifact math (net +1: two added, one removed) is correct.

4. Golden fixture must be re-exported after rebase. tests/fixtures/designer-export/export.json was regenerated at 2.1.3; the byte-exact assertion in t66 needs a fresh export on the rebased tree so it includes the composer agent plus the observability additions.

5. Docs still name monitoring-design.md. Six references the PR removes from core/ but leaves in the shipped docs:

   • docs/reference/03-orchestrator.md:565
   • docs/reference/04-stages/construction.md:450, :515, :538, :550, :1044
AGENTS.md Documentation Policy asks for these to update in the same commit. While there, add rows for observability-requirements.md and observability-design.md wherever the docs enumerate produced artifacts by name.

6. infrastructure-design.md content preservation. Step 5 drops the "Monitoring and Observability" bullet (four areas becomes three) and Step 6 drops the monitoring-design.md bullet. The short "observability tooling" phrase added into Infrastructure Services does not fully preserve the operational-visibility detail (metrics platform, log aggregation, tracing backend, alerting) that lived in the removed bullets. Please either expand the Infrastructure Services phrasing or reference observability-design.md explicitly, so the reader still sees where that content lives.

Non-blocking

7. The PR summary claims (122 -> 124 artifacts, 11 -> 13 agents). The actual diff is 122 -> 123, and agents is untouched by the diff. Worth fixing in the PR body so reviewers do not reconcile silently.

8. core/knowledge/aidlc-quality-agent/nfr-reliability-guide.md is titled "NFR Reliability and Observability Guide", which pre-dates promoting observability to a peer NFR category. Optional pass to reshape the knowledge companions so observability reads like a peer rather than a sub-topic of reliability. Not a merge blocker.

9. Cross-PR interaction with #402 (same author, also touches nfr-design.md + infrastructure-design.md). Whichever merges second will re-apply on top of the first. Heads-up only.

10. Post-rebase, run t66 and confirm the advisory-edge baseline still holds. Adding required consumes changes the dangling calculation for scopes that skip the producer stage; a diff there is a design question, not a mechanical pin update.

Design confirmation (for the record)

The observability chain closes cleanly in scopes that run every stage from nfr-requirements through infrastructure-design. nfr-requirements produces observability-requirements; nfr-design consumes it and produces observability-design; infrastructure-design and observability-setup consume observability-design. Scopes that skip nfr-design do not consume observability-design either, and the dangling-consume machinery from 2.1.7 covers the case where nfr-requirements produces observability-requirements without a downstream consumer in scope.

Ping me post-rebase and I will re-review.