Skip to content

RF — TFW-50: Minimal Agent Commit Attribution

Date: 2026-08-05 Author: Codex / Executor Status: 🟢 RF — Complete; Reviewer evidence pending independent review Parent HL: HL-TFW-50 TS: TS TFW-50

1. What Was Done

New Files

File Description
evidence/EV__TFW-50__minimal_agent_commit_attribution.md Structured source, history, regression, cleanup, and no-publication evidence
RF__TFW-50__minimal_agent_commit_attribution.md Executor result and acceptance trace

Modified Files

File Changes
.tfw/conventions.md Refined the sole normative sentence with exact agent, task, scope, role, and summary derivation; replaced the executor-centric example with the approved Coordinator example
.tfw/glossary.md Refined the concise definition with first-line subject and Git author/committer metadata boundaries without duplicating the grammar
ONB__TFW-50__minimal_agent_commit_attribution.md Recorded the revised RES/TS contract, boundaries, and Coordinator approval
README.md Advanced TFW-50 through revised ONB to RF and linked lifecycle artifacts

Preserved and Verified Files

File Result
.tfw/workflows/handoff.md Byte-identical to 389168a; corrected Step 4 retained
.agent/workflows/tfw-handoff.md Byte-identical to 389168a; unrelated Evidence drift retained
.claude/commands/tfw-handoff.md Byte-identical to 389168a; unrelated Evidence drift retained
RELEASE.md Byte-identical to 389168a; attributed example and explicit push approval retained

2. Key Decisions

  1. Applied RES Iteration 1 C7: one universal conventions owner and one concise glossary reference, with no workflow-wide cue broadcast.
  2. Kept Commit Attribution conditional on commits that occur. No per-stage, WAIT, STOP, workflow, artifact, file, or AC cadence was introduced.
  3. Derived role from conventions §15/Role Lock and scope from open normalized explicit work-slice text; no maintainer/hybrid role or registry was created.
  4. Preserved all four already-correct conflict-reconciliation blobs exactly and audited the wider corpus read-only instead of synchronizing for symmetry.
  5. Reported history as compatibility/searchability evidence only. Explicit TFW-50 prompts and rejected TFW-49 machinery are causal confounders; Reviewer-owned completion remains pending independent review.

3. Acceptance Criteria

  • [x] AC-1: Conventions are the sole normative owner; exact grammar, one sentence, and one example are present; glossary does not duplicate the grammar.
  • [x] AC-2: Exact term derivation and subject-vs-Git-metadata/authentication boundaries are present with no registry, numeric limit, trailer, or body schema.
  • [x] AC-3: The universal rule applies to Coordinator, Researcher, Executor, and Reviewer; corrective added-text cadence scan returned zero.
  • [x] AC-4: Handoff and release reconciliation remains correct; all four preserve blobs are byte-identical to 389168a; unrelated Evidence drift is unchanged.
  • [x] AC-5: Total implementation surface is exactly six existing paths; corrective execution changed only conventions/glossary; wider corpus conflict counts are zero.
  • [ ] AC-6 Reviewer-owned completion: Coordinator, Researcher, and Executor current-task subjects, prior Reviewer compatibility, regression, cleanup, and no-publication state are verified. The current-task Reviewer subject and independent inspection are explicitly DEFERRED to /tfw-review.

4. Verification

  • Static terminology assertions: PASS — normative owner hits 1; glossary grammar duplicates 0; required boundaries present.
  • Corrective diff: PASS — only .tfw/conventions.md and .tfw/glossary.md changed in c7a0055..420fdbe.
  • Total implementation allowlist: PASS — six paths; unexpected 0; missing 0; new implementation files 0.
  • Preserve blobs: PASS — 4/4 equal their 389168a hashes.
  • Wider corpus audit: PASS — commit-family hits 12; push-family hits 6; automatic-push conflicts 0; subject conflicts 0; hybrid subject roles 0.
  • Commit range before EV/RF: PASS — 14 subjects in bc6779e..HEAD; regex-invalid 0.
  • Docs unit tests: PASS — 55 passed in 1.45s.
  • Docs integration/MkDocs build: PASS — 13 passed in 91.22s.
  • Generated-path cleanup: PASS — only newly generated .pytest_cache/, docs/scripts/__pycache__/, and site/ removed; ignored user files preserved.
  • Protected Git state: PASS — hooksPath unset, legacy hook absent, origin/master unchanged, no history rewrite.
  • Build gate: PASS — integration suite completed the real MkDocs build.

5. Evidence

See EV file for evidence details.

Evidence verdict: 0/6 VERIFIED, 1 DEFERRED, 0 BLOCKED, 5 N/A

The DEFERRED item is solely the current-task Reviewer commit and independent inspection; it is not N/A and must be resolved by /tfw-review.

6. Observations (out-of-scope, not modified)

# File Line(s) Type Description
1 .tfw/workflows/docs.md, .agent/workflows/tfw-docs.md, .claude/commands/tfw-docs.md 7 naming Role headers say Coordinator / Reviewer, while conventions §15 assigns docs workflow ownership to Coordinator. No hybrid commit subject exists; cleanup is outside TFW-50.
2 .tfw/workflows/release.md, .agent/workflows/tfw-release.md, .claude/commands/tfw-release.md 7 naming Role headers say Coordinator / Maintainer, while conventions §15 and the active release subject use coordinator. No maintainer subject role exists; cleanup is outside TFW-50.

7. Fact Candidates

# Category Candidate Source Confidence
1 process Commit Attribution must cover every commit-producing TFW role, including Researcher; an Executor/handoff-only consumer model is incorrect. User correction, TFW-50 resumed execution High
2 constraint TFW-50 formats subjects of commits that occur and must not introduce per-stage, per-WAIT, per-STOP, per-workflow, or per-artifact commit cadence. User correction, TFW-50 RES Iteration 1 delegation High

fact-candidates: processed 2026-08-05

8. Strategic Insights (Execution)

# Insight Category Source
S1 Universal applicability through an always-loaded owner can cover all roles without copying cues into every workflow. The implication is that completeness should be measured against behavior and ownership, not symmetric file edits. convention User correction and revised TS/RES
S2 Subject formatting and commit timing are separate policies. Future workflow changes must not infer lifecycle cadence from the existence of Commit Attribution. constraint User correction and revised TS/RES
S3 Prompted compliant history demonstrates compatibility and searchability, not causation or authentication. Verification should always disclose prompt and prior-mechanism confounders. process User evidence boundary, revised TS/RES

9. Diagrams

flowchart TD
    C["Conventions: sole subject-format owner"] --> CO["Coordinator commits"]
    C --> RE["Researcher commits"]
    C --> EX["Executor commits"]
    C --> RV["Reviewer commits"]
    G["Glossary: concise metadata boundary"] -. "points to" .-> C
    H["Handoff cue: preserved"] -. "instantiates" .-> C
    L["Release example: preserved"] -. "instantiates" .-> C
    C -. "does not create" .-> D["Commit cadence"]

RF — TFW-50: Minimal Agent Commit Attribution | 2026-08-05