feat(skills): add deft-interview skill + strengthen swarm/test-with-code rules (#288 #292 #294 #296)

[MScottAdams]

Summary

Add the deft-interview skill and strengthen swarm/test-with-code rules across the framework.

Changes

Task D: skills/deft-interview/SKILL.md (#296, t2.11.1)

• Created skills/deft-interview/SKILL.md with RFC2119 legend and YAML frontmatter
• Deterministic interview loop with 7 rules: one-question-per-turn, numbered options with stated default, other/IDK escape, depth gate, default-acceptance, confirmation gate, structured handoff contract
• Created .agents/skills/deft-interview/SKILL.md thin pointer
• Added AGENTS.md Skill Routing entry for "interview" / "ask questions" / "structured interview"
• Updated deft-setup Phase 1 and Phase 2 to reference deft-interview for Q&A loops
• Added 12 tests to ests/content/test_skills.py

Task A: deft-swarm Phase 6 read-back verification (#288, t1.21.1)

• Added ! rule to Phase 6 Step 1: after resolving rebase conflicts, re-read resolved file and verify structural integrity before git add
• Prefer edit_files over shell regex for CHANGELOG.md and SPECIFICATION.md conflict resolution
• Added 2 anti-patterns

Task B: deft-swarm Phase 6 Slack release announcement (#292, t1.22.1)

• Added Step 6 to Phase 6: generate standard Slack announcement block after GitHub release
• Template includes version, release title, summary, key changes, agent count, duration, PR numbers, release URL

Task C: Strengthen test-with-code rule (#294, t1.23.1)

• AGENTS.md Before committing: new source files MUST include corresponding test files
• main.md Decision Making: distinguish regression coverage vs forward coverage
• deft-swarm Prompt Template CONSTRAINTS: new source files must have tests
• deft-build pre-commit checklist: verify corresponding test file exists

Checklist

• SPECIFICATION.md has task coverage for all changes
• CHANGELOG.md has entries under [Unreleased]
• task check passes (1001 passed, 19 xfailed)
• PR is on a feature branch (not master)
• Conventional commit message with bullet-point body

Closes #288
Closes #292
Closes #294
Closes #296

[greptile-apps]

Greptile Summary

This PR adds the deft-interview skill — a deterministic structured Q&A interview loop with 7 rules — and propagates it into deft-setup, AGENTS.md, and deft-swarm. It also strengthens the test-with-code gate across four surfaces (AGENTS.md, main.md, deft-swarm, deft-build) and hardens deft-swarm Phase 6 with read-back conflict verification and a Slack release announcement step. Sixteen test functions are added to cover the new and updated behaviors.

Confidence Score: 5/5

Safe to merge — all findings are P2 style/coverage suggestions with no correctness or data-integrity impact.

Both remaining findings are P2: a minor UX inconsistency in an RFC2119-legend documentation file (Rule 5 vs Rule 6 acceptance sets), and a missing regression test for a deft-setup integration that is correctly implemented. Neither affects runtime behavior or data integrity. All test-with-code, swarm Phase 6, and deft-interview acceptance criteria are met and tested.

skills/deft-interview/SKILL.md (Rule 5/Rule 6 "ok" inconsistency) and tests/content/test_skills.py (missing deft-setup integration guard)

Important Files Changed

Filename	Overview
skills/deft-interview/SKILL.md	New skill with well-structured 7-rule interview loop; minor UX inconsistency between Rule 5 (accepts "ok" for default acceptance) and Rule 6 (does not include "ok" in the confirmation gate acceptance list)
tests/content/test_skills.py	Adds 16 test functions across sections 30–32 (12 for deft-interview, 2 for Phase 6 read-back, 2 for Slack announcement); no test verifies that deft-setup's Phase 1/Phase 2 actually reference deft-interview
skills/deft-swarm/SKILL.md	Phase 6 read-back verification and Slack announcement step added cleanly; CONSTRAINTS updated with test-with-code rule; anti-patterns updated to match
AGENTS.md	Skill routing entry for deft-interview added; test-with-code "Before committing" rule strengthened correctly
main.md	Decision Making gate now distinguishes regression coverage (a) from forward coverage (b) — clear and well-integrated
skills/deft-build/SKILL.md	Pre-Commit File Review gains item #7 for forward test coverage; language is consistent with other check items
skills/deft-setup/SKILL.md	Phase 1 and Phase 2 now reference deft-interview as the canonical interview loop; pointers are accurate
.agents/skills/deft-interview/SKILL.md	Thin pointer with correct frontmatter and redirect to skills/deft-interview/SKILL.md

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart TD
    A[Calling Skill invokes deft-interview] --> B[Provide: required fields, question defs, optional fields]
    B --> C[Rule 1: Ask ONE question]
    C --> D{User Response}
    D -- Empty/bare enter/default token --> E[Rule 5: Accept Default]
    D -- Explicit answer --> F[Rule 4: Depth Gate Check]
    D -- Other / IDK --> G[Rule 3: Open-ended Follow-up]
    G --> F
    E --> F
    F -- Ambiguity remains --> C
    F -- All required fields captured --> H[Rule 6: Confirmation Gate - Display summary of all answers]
    H --> I{User says yes/confirmed/approve?}
    I -- No --> J[Ask which values to correct - Re-ask specific questions only]
    J --> H
    I -- Yes --> K[Rule 7: Exit with Answers Map JSON]
    K --> L[Calling Skill generates artifact]

Loading

Prompt To Fix All With AI

This is a comment left during a code review.
Path: skills/deft-interview/SKILL.md
Line: 69-77

Comment:
**"ok" accepted at Rule 5 but silent at confirmation gate**

Rule 5 explicitly lists `"ok"` as a valid default-acceptance response, but Rule 6's acceptance list (`yes`, `confirmed`, `approve`) omits it. A user who habitually answers `"ok"` throughout the interview will be silently rejected at the final confirmation gate without an obvious reason — they've been conditioned by Rule 5 that `"ok"` is sufficient.

Consider adding `"ok"` to Rule 6's explicit affirmative list, or adding a clarifying note that the confirmation gate is intentionally stricter.

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

---

This is a comment left during a code review.
Path: tests/content/test_skills.py
Line: 966-1070

Comment:
**Missing regression test for deft-setup → deft-interview integration**

Section 30 tests deft-interview in isolation, but no test verifies that `skills/deft-setup/SKILL.md` Phase 1 and Phase 2 actually reference `deft-interview` — an acceptance criterion listed in `SPECIFICATION.md` (t2.11.1). Currently, the references at lines 107 and 254 of `deft-setup/SKILL.md` could be accidentally removed without failing any test.

A simple guard would be:

```python
def test_deft_setup_phase1_references_deft_interview() -> None:
    """deft-setup Phase 1 must reference deft-interview as the canonical interview loop."""
    text = _read_skill("skills/deft-setup/SKILL.md")
    assert "skills/deft-interview/SKILL.md" in text, (
        "skills/deft-setup/SKILL.md: Phase 1/Phase 2 must reference "
        "skills/deft-interview/SKILL.md (#296, t2.11.1)"
    )
```

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

_{Reviews (3): Last reviewed commit: "fix: address Greptile review findings (b..." | Re-trigger Greptile}

[MScottAdams]

Rebase-only force-push onto updated master (587bafb). No logic changes -- CHANGELOG.md conflict resolved by merging both [Unreleased] entries. Greptile re-review in progress.