idd-all SKILL.md: pseudocode quality (goto bash / branch-exists no-op / stale Failure Modes table)

[kiki830621]

Problem

From verification of #1 (P2 findings 10, 11, 13 — logic+regression reviewers):
Three pseudocode quality issues in idd-all SKILL.md introduced by #1:

1. goto Phase 6 is not valid bash — Phase 5 has:

   if [ "$PATH_AXIS" = "direct-commit" ]; then
     echo "→ direct-commit path: skipping push + PR"
     goto Phase 6
   fi

   Implementers pasting this verbatim get a syntax error.

2. PR-mode branch-already-exists handler is a no-op — Phase 0.5 PR-mode block has:

   if git -C "$CWD" show-ref --verify --quiet "refs/heads/$BRANCH"; then
     # AskUserQuestion: checkout 繼續 or 用 idd/19-...-2?
     :
   fi
   git -C "$CWD" checkout -b "$BRANCH"

   The : is a no-op; the next line guarantees fail with fatal: branch already exists. Failure Modes table promises "AskUserQuestion(checkout / -2 suffix / abort)" but it doesn't exist.

3. Failure Modes table stale — table still references "Phase 0.3 abort" for working-tree-dirty and not-on-default-branch rows, but those gates moved to Phase 0.5 PR-mode-only.

Type

fix

Why these are P2 not P0/P1

None of these block the conceptual contract. (1) is a paste hazard but reading code shows intent. (2) makes branch-already-exists case fall through to a hard fail, which is recoverable. (3) is doc-vs-impl drift, no behavioral consequence.

Recommendation

1. Replace goto Phase 6 with a structural conditional:

   if [ "$PATH_AXIS" = "PR" ]; then
     # ... push + PR creation ...
   fi
   # Unconditional: continue to Phase 6

2. Implement the AskUserQuestion handler properly, OR document that the no-op falls through to a fatal error and that's acceptable. (Pick one.)

3. Update Failure Modes table rows: Phase 0.3 abort → Phase 0.5 abort for the working-tree and default-branch gates.

Source

• Verify FAIL on commit 98b7a71 (idd-all: 增加 HITL（attended）模式 — direct-commit + 允許 sub-skill AskUserQuestion #1)
• Logic reviewer findings 3, 5, 7
• Regression reviewer finding 2
• Master comment: idd-all: 增加 HITL（attended）模式 — direct-commit + 允許 sub-skill AskUserQuestion #1 (comment)

Related: #1

[kiki830621]

Closing Summary

Problem

3 個 SKILL.md pseudocode quality issue (P2 from #1 verify findings 10, 11, 13):

1. Phase 5 用 goto Phase 6 — bash 沒 goto,implementer 照貼會 syntax error
2. PR-mode branch-already-exists handler 是 no-op : 後接必 fail 的 git checkout -b — Failure Modes table 承諾的 AskUserQuestion 不存在
3. Failure Modes table 仍標 "Phase 0.3 abort",但 dirty/branch gate 已搬到 Phase 0.5 — doc-vs-impl drift

Root Cause

SKILL.md 是 markdown 描述 skill behavior,但 bash code blocks 被當成可貼可執行的 reference。寫 invalid bash(goto)或 placeholder no-op 違反這個契約。

Solution

• goto Phase 6 → 結構性 if [ "$PATH_AXIS" = "direct-commit" ]; then ... elif [ "$PATH_AXIS" = "PR" ]; then ... fi,加 comment「both branches fall through to Phase 6 unconditionally below」
• Branch-already-exists handler → 真實的 AskUserQuestion handoff(三選項 a/b/c),exit if 後 agent assigns BRANCH per user choice;包含 fallback else git -C "$CWD" checkout -b "$BRANCH" for 一般 case
• Failure Modes 兩 row 改寫:dirty/not-on-default 標「Phase 0.5 abort (PR mode only)」+ 註明 direct-commit mode falls back to passive warning

Verification

Smoke test 28/28 PASS。Failure Modes 文字已對齊 Phase 0.5 PR-mode-only gates 的實作。

Changes

• a7a4e5e fix: idd-all/idd-issue P2/P3 follow-up batch (Refs idd-all: direct-commit mode 加 passive warning (dirty tree / on default branch) #2 idd-all SKILL.md: pseudocode quality (goto bash / branch-exists no-op / stale Failure Modes table) #3 idd-all Phase 4: ${MAX_ROUND} substitution missing in abort message #4 idd-all Phase 0.5: fork detection 改 fail-closed (gh repo view 失敗時預設 PR mode) #5 idd-all SLUG derivation: BSD/GNU sed alternation (\| not portable) #6 idd-all Phase 5: PR body heredoc 改 quoted (預防未來 RCE) #7 idd-all Phase 0.2: GITHUB_REPO post-derive shape assertion #8 idd-issue Step 2: 強制 sanitize incoming title (length cap / strip control chars) #9 idd-all Examples: 跨 repo invocation 段重複描述,合併 #10)