feat(idd-comment): --type=reply human-facing point-by-point reply (v2.100.0, #269)#271
Merged
Conversation
…c) static check The 23 idd-edit fixtures all invoke the helper directly, so 'SKILL references a variable the helper never emits' bugs shipped with green tests three times (R1 B1 APPEND_BODY, R1 B2 GITHUB_REPO, independently flagged by 3 reviewers). New suite idd-edit-contract embeds a python3 checker: EMITTED derives mechanically from the helper source (emit_assignment calls + TARGETS array — single source, no third variable list), CONSUMED/DEFINED parse the SKILL's fenced bash blocks, and the invariant CONSUMED - DEFINED - EMITTED - ALLOWLIST = empty catches the unsourced-reference class. Allowlist is env only (CLAUDE_PLUGIN_ROOT, IDD_CALLER per the #161 cross-skill contract, ...; GITHUB_REPO deliberately excluded — that WAS bug B2). Checker efficacy is self-tested against a seeded three-violation fixture (detection failing = suite fails), and the 11-name emit surface is frozen (deleting an emit goes red here before it breaks consumers). First live run caught IDD_CALLER and forced its provenance classification — working as intended. SKILL gains a Contract section documenting mechanism + scan boundary (H7 structural class stays with fixture 14). Shape (a) escalation stays trigger-based per the approved plan. GREEN 17/0; sweep 39 suites 0 fail. Refs #163
… (v2.100.0) Seventh comment type for correspondence to a human collaborator (review author, advisor), the recipient-facing counterpart to the six audit-facing types. Per-point structure: verbatim blockquote of the counterpart's point -> what changed and where -> commit/PR/merge-SHA anchor -> honest per-point status. Required --points-from resolves through a three-layer chain (comment URL -> issue-body Original-text blockquote -> user-pasted); verbatim, no paraphrase. verify-before-claim gate (same family as idd-close Step 1.6) keeps unevidenced points open/pending. perspective-writer soft integration with graceful degrade: presence-check present -> calibrate voice/recipient; absent -> print the two install commands and post the anchored draft anyway. No install-time dependency (deliberate opposite of superpowers hard-dependency — calibration is an enhancement, not a canonical-process substitute). Anchoring precedes calibration; calibration must not alter anchored facts. New drift-guard suite idd-comment-reply (19 assertions). run-all-tests: 40 suites, 0 fail. Docs (commands/workflows/README) + CHANGELOG + both manifests bumped to 2.100.0. Spectra change add-idd-comment-reply-type (validate + analyze clean). Refs #269 Related PsychQuant/psychquant-claude-plugins#116, PsychQuant/perspective-writer#1
Verify round 1 (4 lens + Codex + Devil's Advocate) findings, all folded: - DA-1 (HIGH): verify-before-claim was issue-level (git log --grep "#N") but the obligation is per-point. Rewrote R2: the grep is an entry point; the found commit/PR must actually contain the change addressing THAT point, and finding any #N commit must not mark all points resolved. Locked by 2 new drift-guard assertions. - Codex + Regression F1: --points-from "required" vs spec "when unspecified default to issue-body" was a genuine contradiction. Reworded spec so the three layers describe how the REQUIRED value resolves (issue-body is an explicit sentinel), not a default for an absent flag. Dropped --body from reply's 必填 column (Step 2 + spec make it optional). - DA-2 (MEDIUM): bootstrap TaskCreate order (build_comment_body before resolve/verify) contradicted the pipeline "順序是契約". Reordered so resolve_points_source + verify_before_claim precede build_comment_body (= R3 anchoring) and annotated the dependency. - Security F4 + DA-3 (MEDIUM): reply is the only type reproducing third-party verbatim. Added scrub-wins-over-verbatim clause + noted SCRUB_LEVEL is repo-visibility-keyed (own-repo → warn/light, never enforce) so layer-3 pasted external content needs heightened self-review, not tier default. - Test strengthening (Logic LOW-1 / Requirements F3 / Codex gaps): anchored the weak "reply" docs needle, added assertions for layer-1 URL / layer-3 paste / per-point template elements / degrade post-anyway / scrub-wins. Suite 19 -> 27 assertions. - Doc drift: design/proposal run.sh -> test.sh, suite count 39->40. run-all-tests: 40 suites 0 fail. reply suite 27/27. spectra valid. Refs #269
Member
Author
Verify Report — PR #271 (Refs #269)Enginemanual fan-out (4 lens Agents + sequenced Devil's Advocate, model: opus, file-based output) + Codex (gpt-5.6-sol, xhigh, via codex-call) AggregatePASS (after round-1 fix 6 個獨立來源（requirements / logic / security / regression lens + Codex 跨模型盲驗 + Devil's Advocate）全數判 mergeable。無任一 BLOCKING。Round-1 找到 4 個實質不一致（其中 DA-1 是 HIGH、feature 核心價值），已全部在本 PR 修掉並補測試鎖住。 Requirements coverage6/6 spec requirements FULLY addressed（reply per-point 結構 / points-source 三層 / verify-before-claim / perspective-writer soft integration + graceful degrade / anchoring-precedes-calibration / additive + egress）。drift-guard suite 19 → 27 assertions，RED-verified 非 tautological（多個 reviewer 各自剝 needle 驗證會紅）。 Findings（合併去重）
Scope Check
Follow-up Findings Filed (v2.72.0+ #148)
Process notes
|
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.
Summary
idd-comment新增第 7 型--type=reply（v2.100.0）—— 寫給人類 collaborator（review 提出者、指導教授）看的逐點回覆，是六個 audit-facing 型別以外唯一 recipient-facing 的型。處理 #269（after merge 走 /idd-close，見下）（Spectra changeadd-idd-comment-reply-type；validate + analyze clean）。Spec delta（
idd-comment-reply，6 requirements）--points-from；每點 verbatim blockquote → 改了哪/怎麼改 → commit/PR/SHA 錨定 → per-point 狀態；結尾整體狀態行 + metadata marker。git log --grep "#N"/ merged-PR 驗證；無證據寫 open/pending。同族於 idd-close Step 1.6 semantic gate。perspective-writer:perspective-writercalibration（轉交 mention login + target repo correspondence rules 路徑）；缺席 → 印兩行安裝指令照 post。不新增 install-time dependency（刻意與 superpowers hard-dependency 相反：calibration 是 enhancement 非 canonical-process 替代）。Changes
skills/idd-comment/SKILL.md— 型別表 6→7、Step 2 validation 列、#### Template: reply、Reply drafting pipeline（R1–R5）、bootstrap stage tasks、鐵律、範例scripts/tests/idd-comment-reply/test.sh（19 assertions）→ run-all-tests: 40 suites, 0 fail設計背景
D1–D5 於 #269 spectra-discuss 收斂（decision comment）。觸發案例：#141 的六點 review 逐點回覆，當時分兩輪口頭指揮才排出格式——本 PR 把該格式收斂成 first-class 型別。
依賴
PsychQuant/perspective-writerv2.10.0（degrade notice 的安裝指令對公共使用者才成立）Test plan
bash plugins/issue-driven-dev/scripts/tests/idd-comment-reply/test.sh— 19/19 pass（RED→GREEN 已驗證：先寫測試對空 SKILL 全紅，實作後全綠）bash plugins/issue-driven-dev/scripts/run-all-tests.sh— 40 suites 0 failspectra validate/spectra analyze— valid / 0 critical-warning