Skip to content

dev-team: extract upstream v7.7–7.9 (review output discipline, craftsmanship, closed learning loop) + test-after sweep + CI compliance#17

Merged
outofrange-consulting merged 2 commits into
mainfrom
claude/upstream-agentic-dev-team-ox11sn
Jun 25, 2026
Merged

dev-team: extract upstream v7.7–7.9 (review output discipline, craftsmanship, closed learning loop) + test-after sweep + CI compliance#17
outofrange-consulting merged 2 commits into
mainfrom
claude/upstream-agentic-dev-team-ox11sn

Conversation

@outofrange-consulting

Copy link
Copy Markdown
Owner

Extracts what fits omp-dev-team's choices (test-after with refactoring, no TDD, quality first, cost efficiency) from upstream bdfinst/agentic-dev-team v7.7–v7.9, sweeps residual TDD framing, and adds CI to enforce the framework going forward.

Extracted (filtered through omp choices)

  • Deterministic status + finding grouping (v7.9) — new shared knowledge file review-output-discipline.md (status from the highest-severity finding, never volume; enumerate→classify→group into ~3–5 concept-level findings/file). Wired as an anchored reference into all 17 finding-emitting review agents — factored into one shared file rather than upstream's per-agent copy (DRY; propagates determinism + token savings everywhere).
  • Comment hygiene (v7.8)doc-review: tracker-ID refs + orphaned doc comments, capped at suggestion; durable standards (RFC-2119, ISO-4217, CVE) not flagged.
  • Use-the-platform (v7.8)refactor-opportunity-review: reinvented built-ins / helpers / repeated idioms, honoring language and version.
  • Closed learning loop (v7.7)feedback-learning: added the missing "closed" half — a pending-review.jsonl queue + a session-review disposition flow (review keyword). Project-local, plugin-cache-safe, batched (cheap).

Test-after reinforcement + residual-TDD sweep

  • Refactor-after-green promoted to a deliberate per-step pass in the build skill, implementer prompt, and orchestrator Phase 3 (the refactoring half of test-after-with-refactoring).
  • Reframed residual TDD → test-after across triage, qa-engineer, plan, progress-guardian, mutation-testing, quality-gate-pipeline, init-dev-team, test-design-reviewer/advisor, cd-test-architecture, test-strategy, the operating manual, the root README/README.fr ("strict TDD" → forced plan gate), and REVIEW.md (stale tdd-guardspec-guard). Semantics preserved (e.g. mutation-testing fail-on-mutant); bibliographic citations kept factual.

CI: framework-compliance checks (new)

scripts/ci-framework-compliance.mjs (pure Node) + a compliance job automate what prior extraction PRs verified by hand:

  • Anchor resolution — every skill://dev-team-knowledge/<file>.md#<anchor> resolves (file exists; anchor matches a heading slug or a registered index.json anchor). Guards exactly the rename/typo failure mode this work risked across 17 agents.
  • index.json integrity — every keyed file exists.
  • Review-agent wiring — every finding-emitting agent references review-output-discipline.md.
  • Test-after stance — deliberately-removed TDD identifiers can't reappear (allowlist for rationale/historical context).

Currently: 26 anchor refs + 39 index files — 0 violations; negative-tested.

Not extracted (rationale)

  • when-tdd-pays fixtures (v7.7) — re-litigates a settled omp choice (test-after + plan gate).
  • /ship gate ambiguity protocol (v7.7) — wired to a /ship command we don't have; /specs already runs a consistency gate.

Verified

ci-validate-json 23/23 · 10 dev-team extensions compile · unit suite green · framework compliance 0 violations. Bumps dev-team 1.2.0 → 1.3.0. Full rationale in docs/upstream-v7.7-7.9-extraction.md.

🤖 Generated with Claude Code


Generated by Claude Code

claude added 2 commits June 25, 2026 21:00
…rning loop; test-after sweep (upstream v7.7–7.9)

Extract from bdfinst/agentic-dev-team v7.7–v7.9, filtered through omp choices
(test-after with refactoring, no TDD, quality first, cost efficiency):

- v7.9 deterministic status + finding grouping: new shared knowledge file
  review-output-discipline.md (status from highest-severity finding, never
  volume; enumerate→classify→group into ~3–5 concept-level findings/file),
  wired as an anchored reference into all 17 finding-emitting review agents
  (DRY vs upstream's per-agent copy); indexed.
- v7.8 comment hygiene → doc-review (tracker-ID + orphaned doc comments,
  capped at suggestion; durable standards not flagged).
- v7.8 use-the-platform → refactor-opportunity-review (reinvented built-ins,
  helpers, repeated idioms; honor language+version), framing de-TDD'd.
- v7.7 closed learning loop → feedback-learning: pending-review.jsonl queue +
  session-review disposition (review keyword), project-local, cache-safe.

Test-after reinforcement + residual-TDD sweep:
- Refactor-after-green promoted to a deliberate per-step pass in build skill,
  implementer prompt, and orchestrator Phase 3.
- Reframed residual TDD to test-after: triage, qa-engineer, plan, progress-
  guardian, mutation-testing, quality-gate-pipeline, init-dev-team,
  test-design-reviewer/advisor, cd-test-architecture, test-strategy,
  operating-manual; root README/README.fr (strict TDD → forced plan gate);
  REVIEW.md (stale tdd-guard → spec-guard).

Not extracted: when-tdd-pays fixtures (re-litigates a settled omp choice);
/ship gate ambiguity protocol (no /ship command here). Bibliographic TDD
citations kept as factual.

Verified: ci-validate-json 23/23; 10 extensions compile; unit suite green;
both anchors resolve from all 17 agents. Bumps dev-team 1.2.0 → 1.3.0.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AKLSTtwu9gVqFrivndRAsF
…tegrity, review wiring, test-after guard)

New scripts/ci-framework-compliance.mjs (pure Node, no deps) + a `compliance`
CI job automate what prior extraction PRs verified by hand:

- Anchor resolution: every skill://dev-team-knowledge/<file>.md#<anchor>
  reference resolves — file exists and anchor matches a heading slug or a
  registered index.json anchor (the project's hand-authored anchor registry).
  Guards exactly the rename/typo failure mode the 17-agent wiring risked.
- index.json integrity: every keyed file exists.
- Review-agent wiring: every finding-emitting agent references
  review-output-discipline.md (allowlist: progress-guardian).
- Test-after stance: deliberately-removed TDD identifiers (tdd-first, tdd-guard,
  test-driven-development, RED-GREEN) can't reappear, outside a rationale/
  historical allowlist.

Currently: 26 anchor refs + 39 index files — 0 violations. Negative-tested
(broken ref + TDD token both caught).

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AKLSTtwu9gVqFrivndRAsF
@outofrange-consulting outofrange-consulting merged commit bb65fb5 into main Jun 25, 2026
18 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants