fix(workflow): semantic accuracy check in mandatory pre-commit file review (#274)

[MScottAdams]

Summary

Add semantic accuracy check to the mandatory pre-commit file review step in both deft-roadmap-refresh and deft-build skills.

Closes #274

Changes

• skills/deft-roadmap-refresh/SKILL.md: Added semantic accuracy check as a fourth bullet under Phase 4 pre-flight mandatory file review -- verify that counts, claims, and summaries in CHANGELOG entries and ROADMAP changelog lines match the actual data in the commit
• skills/deft-build/SKILL.md: Added equivalent semantic accuracy check as item 4 in the Pre-Commit File Review numbered list, renumbering existing semantic contradictions (now 5) and strength duplicates (now 6) checks
• CHANGELOG.md: Added entry under [Unreleased] ### Fixed for t1.15.1

Spec Task

t1.15.1 (#274) -- acceptance criteria:

• skills/deft-roadmap-refresh/SKILL.md Phase 4 pre-flight mandatory file review step contains semantic accuracy check
• skills/deft-build/SKILL.md pre-commit checklist contains equivalent semantic accuracy check

Checklist

• SPECIFICATION.md has task coverage (t1.15.1)
• CHANGELOG.md entry under [Unreleased]
• task check passes
• /deft:change N/A (fewer than 3 files changed)

[greptile-apps]

Greptile Summary

This PR adds a semantic accuracy check to the mandatory pre-commit file review in both skills/deft-build/SKILL.md (as new item 4 in the numbered list, renumbering previous items 4→5 and 5→6) and skills/deft-roadmap-refresh/SKILL.md (as a fourth bullet under the Phase 4 mandatory file review), plus a corresponding [Unreleased] CHANGELOG entry. The changes are focused, correctly scoped to the acceptance criteria, and the two skill files are internally consistent.

Confidence Score: 5/5

Safe to merge — changes are additive documentation edits with no runtime risk.

All remaining findings are P2 (style/clarity). The only issue is that the illustrative example in deft-build is borrowed from the roadmap-refresh domain, which is a minor clarity gap and does not affect correctness or enforcement. Both acceptance criteria from the spec task are satisfied, the renumbering in deft-build is correct, and the CHANGELOG entry is accurate.

No files require special attention; the P2 note on skills/deft-build/SKILL.md line 142 is a minor wording improvement, not a blocker.

Important Files Changed

Filename	Overview
skills/deft-build/SKILL.md	Adds semantic accuracy as item 4, renumbers old 4→5 and 5→6; the illustrative example is borrowed verbatim from the roadmap-refresh domain and doesn't align with build-workflow semantics
skills/deft-roadmap-refresh/SKILL.md	Adds semantic accuracy as a fourth bullet in Phase 4 mandatory file review; wording and example are domain-appropriate and consistent with the existing bullet style
CHANGELOG.md	Adds [Unreleased] Fixed entry for #274/t1.15.1 describing the new semantic accuracy check; entry is correctly placed and accurately describes the change

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart TD
    A[Pre-Commit / Pre-Flight] --> B[Re-read ALL modified files]
    B --> C{Check 1: Encoding errors}
    C --> D{Check 2: Unintended duplication}
    D --> E{Check 3: Structural issues}
    E --> F{Check 4: Semantic accuracy NEW}
    F --> G[deft-build only]
    G --> H{Check 5: Semantic contradictions}
    H --> I{Check 6: Strength duplicates}
    F --> J[deft-roadmap-refresh only]
    J --> K[Proceed to commit / push]
    I --> K

Loading

Prompt To Fix All With AI

This is a comment left during a code review.
Path: skills/deft-build/SKILL.md
Line: 142

Comment:
**Example is roadmap-refresh-specific**

The parenthetical example `"triaged 4 issues"` is copied verbatim from the roadmap-refresh skill and has no meaning in a build workflow. A reader of `deft-build` who has never used `deft-roadmap-refresh` won't recognise this as relevant, which reduces the check's signal value for the intended audience.

Consider replacing with a build-domain example such as counts of tasks implemented, tests written, or phases completed:

```suggestion
4. ! **Semantic accuracy** -- verify that counts, claims, and summaries in CHANGELOG entries match the actual data in the commit (e.g. "added 3 tests" must match the tests actually added, task IDs cited must match the tasks actually implemented)
```

How can I resolve this? If you propose a fix, please make it concise.

_{Reviews (3): Last reviewed commit: "fix(workflow): add semantic accuracy che..." | Re-trigger Greptile}

[MScottAdams]

Rebase-only force-push onto updated master (PR #284 merged). No logic changes -- rebased to resolve CHANGELOG.md insertion-point conflict.