chain (multi-root): 3 issues — idd-close: Residue acknowledgement prompt — close the F3 write-only loop#110
Merged
kiki830621 merged 7 commits intoMay 20, 2026
Conversation
#103 PR #104 expanded idd-issue Step 5 from metadata-only to also echo the AI-rendered ## Type / ## Expected / ## Actual + plain-language interpretation. For human readers this is strictly better — misparse becomes catchable from terminal. For CI / /loop scripts that hard-parse Step 5 stdout, this is a silent surface change. Adds one ⚠ paragraph mirroring the v2.55.0 --no-multi-finding CI warning precedent at line 994. Verbatim wording from #107 issue body, scoped to Step 5 (same anchor as the v2.64.0 echo paragraph this warning is about). No behavior change — purely documentation.
…ing (Refs #108) 5-template + 3-satellite consistency family — this commit covers the 4 PR-body wording sites that #102 left out of scope: - skills/idd-implement/SKILL.md:503 (PR template Pending review line) - skills/idd-all/SKILL.md:755 (Phase 5 PR-body Pending review line) - references/pr-flow.md:135 (canonical Pending review line) - references/chain-flow.md:254 (cluster variant) All swap legacy 'Pending: human review of this PR + /idd-close after merge' to verify-gated default: 'Verify-gated: verify PASS — ready to merge → /idd-close ... after merge'. Per Option A from #108 diagnosis: idd-implement does NOT accept --review flag (--review remains idd-all + idd-all-chain only); direct idd-implement invocations get the default wording without conditional. Remaining #108 work in subsequent commits: F3 (idd-all-chain Phase 4 stdout --review-aware), Trace 1 example refresh (idd-all 893-901), DA3 wording precision (orchestrator-scope messaging-only).
…efs #108 F3) Per PR #109 verify regression finding F3: Phase 4 final report stdout text (the verify-gated terminal disposition line + Next steps block) was NOT --review-aware, while the PR-body checklist already was. Build VERIFY_LINE + NEXT_STEPS as separate vars via explicit if/else BEFORE the heredoc (NOT inline ${VAR:-word} — that mutex pitfall already caught in PR #109 F1, hardcoded into the doctrine). Default: 'Verify: verify-gated PASS across cluster — cluster ready to merge' + 'Next: 1. Merge ... 2. /idd-close ...' --review: 'Verify: verify-gated PASS across cluster — awaiting human acceptance (re-opened confirmation loop per --review)' + 'Next: 1. Review PR ... 2. Merge after acceptance 3. /idd-close ...' This closes the F3 satellite of #108's 5-template + 3-satellite scope.
…efs #108) Two remaining satellites from #108's 5-template + 3-satellite scope: **Trace 1 example refresh** (idd-all/SKILL.md 893-901, idd-all 鐵律 line 843): - Default example now shows 'Verify: verify-gated PASS' + 'Next: merge' - Adds parallel --review variant block showing 'awaiting human acceptance (re-opened confirmation loop per --review)' + 'merge after acceptance' - 鐵律 line: 'messaging-only' → 'orchestrator-scope messaging-only' with explicit #108 DA3 citation **DA3 wording precision** (3 sites + MANIFESTO): - idd-all/SKILL.md:117 args parsing comment → 'orchestrator-scope messaging-only effect' - idd-all-chain/SKILL.md:118 args parsing comment → same swap - MANIFESTO.md Human-in-the-loop --review paragraph → 'orchestrator-scope' Rationale: '--review is messaging-only' is true AT THE ORCHESTRATOR (does NOT change skill behavior, does NOT make orchestrator wait), but humans + CI parsers downstream can react to the changed text differently — so it isn't messaging-only in the strict end-to-end sense. The qualifier prevents that misreading. Closes the F3 + Trace 1 + DA3 satellites of #108's scope. Remaining for #108: CHANGELOG + plugin.json bump (next commit).
…ly loop (Refs #105) Per #103 PR #104 verify Devil's Advocate D2 finding: `### Residue` shipped in #103 as a producer-only field — idd-implement / idd-verify / idd-close / idd-list / idd-update / idd-all all read OTHER Diagnosis sections but none read Residue. Result: latent ritual-drift risk — without a downstream consumer pressuring the field to stay honest, it becomes ignored boilerplate over time. #105 closes the loop by giving Residue its first downstream consumer at close time: - NEW Step 3.6 between Step 3.5 (closing follow-up keyword scan) and Step 4 (publish + close), mirroring Step 3.5's drafted-summary-scan pattern. - Reads the **latest** `## Diagnosis` comment's `### Residue` section (mirrors Step 0 supersession — older drafts are deliberation history, not current contract). - Trigger logic: silent skip when no Diagnosis exists (legacy), no `### Residue` section (pre-v2.64.0 format), or content is exactly `(none)` (the explicit empty-state marker the template requires). - When triggered: AskUserQuestion 3-option per IC_R011 canonical pattern — `still residue`, `file as follow-up issue(s)`, `skip`. - Audit trail PATCH per canonical heading-conventions table — every branch records something to the closing summary (5 outcomes covered: acknowledge, file, skip, silent, env-var rollback). - Manifest helper integration: filed follow-ups get spawn manifest entries when chain context active (additive, silent skip otherwise). - Step 0.5 Bootstrap Task List adds `residue_acknowledgement` entry alongside `closing_followup_keyword_scan` for visible per-step accountability. Non-blocking (SHOULD-tier) per closure-tier IC_R011 eligibility: closure is mostly mechanical action with text artifact; hard-blocking on every residue declaration would multiply close-time friction. The value is making 'residue is real, what happened to it?' a visible deliberation moment. Placement rationale: Step 3.5 already established the 'scan drafted closing summary + AskUserQuestion before publish' pattern. Step 3.6 mirrors it — same draft/scan/PATCH flow, different source (Residue vs orphan keyword mentions). Both must run BEFORE Step 4 so the audit PATCH happens to the same in-memory draft.
Cluster release covering 3 chain-solved issues on branch idd/chain-multi-5439e76b-idd-close-residue-acknowledgement-prompt: - #105 (Plan-tier): idd-close Step 3.6 Residue Acknowledgement — closes the F3 write-only loop from #103. - #107 (Simple): idd-issue Step 5 ⚠ paragraph for CI/loop hard-parse warning. - #108 (Plan-tier): 5-template + 3-satellite consistency family — 4 PR-body wording syncs + idd-all-chain Phase 4 stdout --review dispatch + idd-all Trace 1 example refresh + DA3 wording precision. Minor bump justification: new idd-close step (additive behavior at close time) + 4 default PR-body wording strings change (user-visible diff in every PR opened by IDD orchestrators) + new conditional in idd-all-chain Phase 4 stdout. Patch would have under-claimed. Marketplace.json sync deferred to /idd-close Step 6.5 chain per repo precedent. CHANGELOG entry covers all 3 issues.
Verify findings dispatched in same PR per feedback_verify_fix_same_pr: - H1 (regression): openspec/specs/idd-all-chain/spec.md:161 contract drift — updated section heading + bullet wording to match shipped chain-flow.md (## Review status + Verify-gated default + --review opt-in branch). Otherwise live spec contracts pre-v2.66.0 wording that v2.66.0 removes from emission. - L3 (regression): chain-flow.md:249 section heading swap '## Pending review' → '## Review status'. Was missed when idd-all-chain SKILL.md got the cosmetic heading swap; chain-flow.md is the documentation contract and should match. - L1 (regression): idd-issue/SKILL.md:880 parenthetical pointed at 'line 994' but actual reference moved to line 996 after paragraph insertion (off-by-2 from the inserted paragraph itself). Drop the line-number reference entirely — paths rot, the descriptive '--no-multi-finding CI warning below' is enough. - Codex finding (nit): CHANGELOG.md #107 entry claimed 'Verbatim wording from the issue body' but shipped wording adds two skill- internal items (#107 self-reference + precedent parenthetical). Softened to 'near-verbatim ... with two skill-internal additions' to keep CHANGELOG accuracy honest. Refs #105 #107 #108
Contributor
Author
/idd-verify --pr 110 — cluster verify reportPhase: verified (verify-gated PASS)
Aggregate verdict: PASS (post-fix)
Findings dispatched in-PR (per
|
| # | Severity | Source | What | Where |
|---|---|---|---|---|
| H1 | HIGH | regression | Live spec contract drift — openspec/specs/idd-all-chain/spec.md:161 contracted old Pending: human review wording that v2.66.0 removes from emission |
Updated section heading + bullet wording to match shipped chain-flow.md |
| L3 | LOW | regression | Heading dissonance — chain-flow.md:249 ## Pending review but content was Verify-gated checkbox |
## Pending review → ## Review status |
| L1 | LOW | regression | Off-by-2 line number — idd-issue/SKILL.md:880 ⚠ paragraph cited "below at line 994" but actual reference moved to 996 after paragraph insertion |
Dropped specific line number — paths rot |
| Codex nit | LOW | Codex | CHANGELOG #107 entry claimed "Verbatim wording from the issue body" but shipped adds #107 self-reference + parenthetical |
Softened to "near-verbatim ... with two skill-internal additions" |
Per-issue verdict
#105 — idd-close Residue Acknowledgement: PASS
- Step 3.6 inserted at correct position (line 405, between Step 3.5 at 337 and Step 4 at 499)
awkextraction:/^### Residue/{flag=1; next} flag && /^###? /{flag=0} flag— stops at H2/H3 correctly, EOF terminates scan- 3-branch silent-skip (missing /
(none)/ pre-v2.64.0) - AskUserQuestion 3-option per IC_R011 canonical (file all / file selected / skip)
- 5-row audit-PATCH table (acknowledge / file all / file selected / skip / silent-skip)
- Step 0.5 Bootstrap
residue_acknowledgementtask entry at line 166 manifest-append.sh9-arg invocation matches idd-implement Step 5.7 precedent (ROOT_ID_FOR_MANIFEST="${IDD_CHAIN_CURRENT_ROOT_ID:-${NUMBER:-}}")- Security:
RESIDUE_CONTENTinterpolated through quoted variable expansion across all 7 use sites; noeval/bash -c;head -c 4000cap mitigates DoS - Rollback hatch present (
IDD_LOW_BAR_RESIDUE_ACK=false)
#107 — idd-issue Step 5 CI/loop hard-parse warning: PASS
- ⚠ paragraph at exact anchor — line 880, between v2.64.0 echo block (878) and
提示下一步(882) - Wording near-verbatim per issue body
## Expected; skill-internal#107self-reference + precedent parenthetical preserved (CHANGELOG softened to match — fix commit) - Mirrors v2.55.0
--no-multi-findingCI warning precedent shape - Off-by-2 line-number reference dropped (fix commit)
#108 — 5-template + 3-satellite consistency: PASS
- 4 PR-body templates synced (idd-implement:503, idd-all:755, pr-flow:135, chain-flow:254)
- F3 — explicit if/else built BEFORE heredoc (idd-all-chain Phase 4 stdout, lines 707-727) — avoids
${VAR:-word}mutex from PR [enhancement] acceptance-review 缺在 IDD mode model — 「人不 review」需要 first-class 做法 #109 F1. Empirically traces correctly for bothREVIEW_FLAG=""andREVIEW_FLAG="--review" - Trace 1 example refresh with parallel default +
--reviewvariant blocks - DA3 wording —
orchestrator-scope messaging-onlyconsistent in idd-all (lines 117 + 843), idd-all-chain (line 118), MANIFESTO (line 119) - Option A respected —
grep -- "--review" skills/idd-implement/SKILL.mdreturns 0 hits (CLI argument-hint clean) - chain-flow.md heading swap (fix commit) syncs to idd-all-chain SKILL.md heading swap shipped pre-fix
- Live spec contract (fix commit) updates
openspec/specs/idd-all-chain/spec.md:161to match emission
Shared
plugin.json2.65.0 → 2.66.0 (minor — justified by new step + 4 wording + new conditional)CHANGELOG.md[2.66.0]entry covers all 3 issues;### Noteshonest about marketplace.json deferral
Scope discipline
No cross-issue creep verified:
#105→skills/idd-close/SKILL.mdonly#107→skills/idd-issue/SKILL.mdonly#108→ 4 PR-body sites + idd-all-chain (Phase 4 stdout + DA3 args comment + heading swap) + idd-all (Trace 1 + 鐵律 + DA3 args comment) + MANIFESTO (DA3 wording) + pr-flow + chain-flow + spec- Shared: CHANGELOG + plugin.json
Devil's Advocate dismissal note
DA's findings file claims all 4 sibling reviewers reviewed a different PR (Swift MailController.escapeForAppleScript dedup). Verified false: actual sibling findings files (timestamps 2026-05-20 10:35-10:38) all correctly review PR #110 (cite idd-close Step 3.6, idd-issue Step 5, F3 if/else, MANIFESTO line 119). Most likely cause: DA polling succeeded earlier when stale 2026-05-19 sibling files (since overwritten) were on disk. DA's 10 adversarial angles, applied to the actual PR #110 diff (independent of sibling files), surface no new blocking findings — only deferred-tech-debt observations already documented as Residue in the diagnoses.
Next: merge ready
Cluster ready to merge per verify-gated doctrine:
- Merge chain (multi-root): 3 issues — idd-close: Residue acknowledgement prompt — close the F3 write-only loop #110 (squash recommended — single review surface)
/idd-close #105 #107 #108(per-issue closing summary required, no shortcut)- Distribution sync to v2.66.0 via
/plugin-tools:plugin-update issue-driven-dev
Verify: 6-AI ensemble PASS (post-fix). Findings dispatched in same PR per feedback_verify_fix_same_pr.
This was referenced May 20, 2026
kiki830621
added a commit
that referenced
this pull request
May 20, 2026
…n-dev v2.66.0 Distribution sync after PR #110 merge (cb19802) closing #105 #107 #108: - .claude-plugin/marketplace.json: 2.65.0 → 2.66.0 + full description chain (NSQL doctrine follow-up family rationale + 4 deliverables) - plugins/issue-driven-dev/README.md: Version History row v2.56.0–v2.65.0 → v2.56.0–v2.66.0, appended v2.66.0 paragraph with per-issue scope + 6-AI verify outcome + fix commit reference Refs #105 #107 #108
kiki830621
added a commit
that referenced
this pull request
May 20, 2026
… family from #48 verify (#113) chain (multi-root): 5 issues — idd-issue multi-finding spec hardening family from #48 verify (#113) Cluster PR closes 5 sister bugs from #48's 6-AI verify, all same-file (skills/idd-issue/SKILL.md multi-finding mode), shipped as one chain: * #75 (security) — Content sanitization contract: dual-track (jsonl verbatim per IC_R007 + GitHub body sanitized via Unicode-aware Python filter covering C0/DEL/C1/bidi-override U+202A-U+202E + U+2066-U+2069 Trojan-Source CVE-2021-42574 family); sanitize_source_label() refuse-on-@-token cross-references rules/tagging-collaborators.md; jq --arg / --argjson parameter binding mandate. * #76 (bug) — run_id collision + symlink hardening: ISO-8601 millisecond precision dispatch chain (GNU %3N → Python datetime fallback for BSD date hazard on macOS → .000Z last resort); TOCTOU symlink check fail-closed; JSONL_WRITE_GUARD invoked at run-start to lock canonical path before Stage 4 footer composition (prevents nonce-retry path mutation invalidating footer URLs). * #77 (enhancement) — 7 corner-case spec contract gaps closed: flag-conflict refusal layering, partner_eligible_set formal definition, Stage 3 Edit-row soft cap >5, [Back to top-3] 5th Other option, Stage 1 path canonicalization, agent-crash known gap, Stage 4.5 unattended fallback. * #79 (enhancement) — Audit trail completeness: abort-path now writes minimal aborted: true jsonl with partial timestamps (footer URLs valid post-abort); > **Action** footer line; "srt" source_type enum. * #80 (enhancement) — Stage 1 + Stage 2 LLM determinism: anchor heuristics for MAY merge/split clauses; max_possible_score formula explicit; N<3 degenerate-case picker shape table. Plus: CHANGELOG v2.67.0 + plugin.json 2.66.0 → 2.67.0 (minor — additive spec hardening). Verify dispatched 4 HIGH blocking findings in-PR (commit 0fe06ed): F75-1 (CJK UTF-8 corruption via byte-level tr), F76-1 (BSD date %3N silent literal output), F76-2 (JSONL_WRITE_GUARD dead code + RANDOM 15-bit comment-code drift), #79 spec/impl contradiction. 5/6-AI verify PASS post-fix (Codex hung at 12+ min — process gap recorded per v2.59.0+ convention, precedent: PR #110 DA stale-file). Refs #75 #76 #77 #79 #80
4 tasks
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Refs #105 #107 #108
Summary
Multi-root chain (N=3 roots: 105 107 108) solved as one cluster via
/idd-all-chain(v2.60+, traversal=dfs). All 3 are NSQL doctrine consistency follow-ups from #103 (PR #104) + #102 (PR #109), grouped together because they share the theme "follow up the IDD doctrine work":idd-closeStep 3.6 Residue Acknowledgement: closes the F3 write-only loop from IDD human-in-the-loop should conform to the NSQL confirmation protocol (v4.1.0) #103 by giving### Residuea downstream consumer at close time.idd-issueStep 5 CI/loop hard-parse warning: defensive doc mirroring v2.55.0--no-multi-findingprecedent.--reviewdispatch + idd-all Trace 1 example refresh + DA3 wording precision.Cluster overview
c9cfa5951373fc7f94c96+7a81617+37083040370516Per-issue details
#105 (root_id=105) — idd-close: Residue acknowledgement prompt — close the F3 write-only loop
See issue #105 for Diagnosis (Plan-tier, Layer V V1=2 V4=3 untriggered) + Implementation Complete.
NEW Step 3.6 between Step 3.5 (closing follow-up keyword scan) and Step 4 (publish+close) in
idd-close. Reads latest## Diagnosiscomment's### Residuesection, silent skips when content is(none)or section missing, AskUserQuestion 3-option when non-empty. SHOULD-tier (non-blocking) per IC_R011 closure-tier eligibility.#107 (root_id=107) — idd-issue Step 5: CI/loop hard-parse warning
See issue #107 for Diagnosis (Simple, Layer 1 disqualifier — prose update) + Implementation Complete.
One ⚠ paragraph addition at
skills/idd-issue/SKILL.mdStep 5, anchored after the v2.64.0 echo paragraph (per #103). Verbatim wording from issue body's## Expected. Mirrors v2.55.0--no-multi-findingCI warning precedent at line 994.#108 (root_id=108) — Sync 5-template + 3-satellite PR-body wording to NSQL doctrine
See issue #108 for Diagnosis (Plan-tier, Layer V V1=2 V4=3 untriggered) + Implementation Complete.
Wording sync across:
idd-implementStep 5.5,idd-allPhase 5,pr-flow.md,chain-flow.md)idd-all-chainPhase 4 final stdout--review-aware (built via explicitif/elseto avoid the${VAR:-word}mutex pitfall from PR [enhancement] acceptance-review 缺在 IDD mode model — 「人不 review」需要 first-class 做法 #109 F1)idd-allexample block refreshmessaging-only→orchestrator-scope messaging-onlyin 3 sitesReview status
/idd-verify --pr <this PR>)/idd-close #105 #107 #108per issue after merge🤖 Generated by /idd-all-chain. Do NOT add GitHub close trailers (Closes/Fixes/Resolves) — IDD discipline requires manual /idd-close per issue after merge to enforce checklist gate + per-issue closing summary.