fix(planning): expose per-question answered flag in /api/state

[kyle-sexton]

Closes #5459

Summary

/api/state reported state: open for a question that already had a terminal decision, because state is dependency staleness (open, stale, upstream-pending, archived), not an answered flag. A consumer could not tell answered from unanswered questions.

Fix

Hub.state() in plugins/planning/surface/server.py now sets answered on each question, computed with the page's own rule (exporters.latest_decision). README.md and schema/questions.schema.json document answered and state that state is not an answered flag. planning bumped to 0.47.0 with a changelog entry. Finding 15a (page sync after record-terminal) did not reproduce in the issue's repro pass, so no code change there.

Verification

• python3 -m unittest test_server test_schema test_exporters test_round in plugins/planning/surface: 330 tests OK (1 skipped), including a new test for answered.
• scripts/check-changelog-parity.sh --check --check-order: pass.
• scripts/validate-plugins.sh: all manifests and the catalog validated.

Related

Refs #5191 (finding 15 origin).

🤖 Generated with Claude Code

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-09-29T23:05:59.658268Z	5d44811	Draft marked ready

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

[claude]

Claude encountered an error after 0s —— View job

---

I'll analyze this and get back to you.

[claude]

Claude encountered an error after 0s —— View job

---

I'll analyze this and get back to you.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 5d44811e49

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[claude]

Claude finished @kyle-sexton's task in 2m 31s —— View job

---

Code review complete

• Gather context (PR diff, changed files, prior review comments)
• Review server.py changes (answered flag logic)
• Review schema/README/changelog updates
• Review new test coverage
• Post findings

No blocking findings.

Notes

• Codex's earlier P1 (reopen reading answered: true) is already fixed on this head: state() now checks the record's decision (bool((latest_decision(...) or {}).get("decision"))) instead of just presence of a record (server.py#L807-L812), and TestAnswered.test_4_reopen_is_not_answered (test_server.py#L1382-L1387) covers accept-then-reopen. I traced decision_view/newest_decision in exporters.py and confirmed a live reopen yields decision: None with a real updatedAt, so it still wins the "newest" comparison and correctly reads as unanswered.
• The local from exporters import latest_decision inside state() (server.py#L807) is necessary, not a style slip — exporters.py imports from server at module top, so a top-level import the other way would cycle. The comment correctly documents why.
• schema/questions.schema.json, README.md, and CHANGELOG.md accurately describe the new answered semantics and match the implementation; plugin.json's 0.48.0 bump is consistent with the changelog entry (the PR description's "bumped to 0.47.0" is just a stale line in the write-up, not a code issue).
• Test coverage (TestAnswered, 4 cases) exercises the real paths (record-terminal CLI and the page /api/state POST), not just unit-level stubs, and matches the state-vs-answered distinction the PR is about.

Unable to verify: this lane's sandbox denied git fetch, scripts/run-ruff.sh, and python3 -m unittest here, so I could not independently re-run the lint/test commands the PR author reports; the analysis above is from static reading of the diff and the referenced exporters.py helpers only.

[claude]

Claude finished @kyle-sexton's task in 1m 4s —— View job

---

Security review (/review:security-review)

• Gather context (PR body, review comments, changed files)
• Invoke /review:security-review per the custom instructions
• Post findings

Result: skip gate hit — no security-relevant surface.

I reviewed the full diff (gh pr diff 5483) against this skill's criteria. The change is confined to computing and exposing one additional boolean field, answered, on /api/state:

• plugins/planning/surface/server.py — Hub.state() now calls exporters.latest_decision(x, r.get("responses", {})) on already-loaded local JSON (questions.json / responses.json) and sets x["answered"] from the result. No new input parsing, no new endpoint, no change to authn/authz, no change to the token-guarded POST /api/answer path, no injection surface (command/SQL/path/template) touched.
• plugins/planning/surface/exporters.py:457-470 (newest_decision/latest_decision, unchanged by this PR) is pure data selection over the same in-memory structures the endpoint already loads.
• README, schema, changelog, and manifest-version edits are documentation/metadata only.
• test_server.py adds test coverage only.
• The server remains 127.0.0.1-only per its module docstring, unchanged by this PR.

This is skip-gate criterion 2 (change has no security-relevant surface after reading the diff): a pure read-time computation over data the handler already had, with no new trust boundary, no new attacker-controlled input, and no widening of what any caller can do. No instruction-surface deletions (CLAUDE.md/AGENTS.md/rules/skills/hooks) appear in this diff either, so that lens doesn't apply.

GitHub Actions/workflow hardening is out of scope for this lane (zizmor's) — not applicable here regardless, since no workflow files changed.

No findings. Nothing posted as inline comments.

[claude]

Claude finished @kyle-sexton's task in 1m 0s —— View job

---

Security review (/review:security-review)

• Gather context (PR body, review comments, changed files)
• Invoke /review:security-review per the custom instructions
• Post findings

Result: skip gate hit — no security-relevant surface.

I reviewed the full diff (gh pr diff 5483, head 232969a) against this skill's criteria. The change is confined to computing and exposing one additional boolean field, answered, on the existing /api/state GET response:

• plugins/planning/surface/server.py#L807-L816 — Hub.state() calls exporters.latest_decision(x, r.get("responses", {})) on questions.json/responses.json, both already loaded locally by this handler, and sets x["answered"] from bool((... or {}).get("decision")). No new input parsing, no new endpoint, no change to /api/state's (lack of) auth requirement, no change to the token-guarded POST /api/answer/POST /api/lease paths, no injection surface (command/SQL/path/template) touched.
• latest_decision/newest_decision in exporters.py are unchanged by this PR — pure data selection over the same in-memory structures the endpoint already had.
• README (README.md#L71), schema/questions.schema.json, CHANGELOG.md, and the plugin.json version bump are documentation/metadata only.
• test_server.py adds test coverage only (TestAnswered, 4 cases, including the reopen case from Codex's earlier P1, which is fixed on this head — answered now checks the record's decision value, not just record presence).
• The server remains 127.0.0.1-only, per the module docstring (server.py#L1) and the README's Security model section (README.md#L75), unchanged by this PR.

This is skip-gate criterion 2: a pure read-time computation over data the handler already loaded, with no new trust boundary, no new attacker-controlled input, and no widening of what any caller of /api/state can already do. No instruction-surface deletions (CLAUDE.md/AGENTS.md/rules/skills/hooks) appear in this diff, so that lens doesn't apply either.

GitHub Actions/workflow hardening is out of scope for this lane — not applicable regardless, since no workflow files changed in this PR.

No findings. Nothing posted as inline comments.

[claude]

Claude finished @kyle-sexton's task in 2m 54s —— View job

---

Code review (/review:code-review) — PR #5483

• Gather context (PR diff, changed files, prior review comments)
• Invoke /review:code-review and follow its instructions
• Post findings

No blocking findings.

What I checked

• Traced state() in server.py#L800-L816: answered is now bool((latest_decision(x, r.get("responses", {})) or {}).get("decision")). This correctly resolves the earlier Codex P1 (reopen reading answered: true): decision_view() in server.py sets "decision": None for a reopen event, so bool(...) on that record is False.
• Cross-checked record-terminal (round.py op_record_terminal, around line 584): a.decision is restricted to {"accept", "alt", "own", "defer"} (the label dict there only maps those four), all truthy strings — so bool(...get("decision")) can never spuriously read a real terminal decision as unanswered.
• TestAnswered in test_server.py (test_1..test_4) exercises unanswered, terminal-decision, page-accept, and accept-then-reopen — the accept-then-reopen case (test_4_reopen_is_not_answered) is a real regression test for the fixed P1, not just a stub.
• README.md, schema/questions.schema.json, and CHANGELOG.md describe the same answered semantics as the code (true when a page or terminal decision counts), and the schema addition is purely additive (new optional boolean property) — no risk to existing questions.json validation.
• plugin.json is at 0.48.0, consistent with the changelog's top entry; the PR description's "bumped to 0.47.0" is a stale line in the write-up only, not a code discrepancy (already flagged by a prior review pass on this PR).

Unable to verify: this sandbox denies python3 -m unittest and git fetch, so I could not independently re-run the author-claimed test suite (330 tests OK) or the changelog-parity/validate-plugins scripts — the analysis above is from static reading of the diff and the referenced exporters.py/round.py helpers only.

This PR has already been through two review passes on this branch (general code review + security review), both of which found the Codex P1 already fixed on this head. My pass, done independently against the current HEAD SHA, agrees: no new findings.

[claude]

Claude finished @kyle-sexton's task in 1m 9s —— View job

---

Code review (/review:code-review) — PR #5483

• Gather context (PR diff, changed files, prior review comments)
• Invoke /review:code-review and follow its instructions
• Post findings

Skip gate hit — criterion 4 (a successful review from this lane already applies to this head).

I diffed the current HEAD (26e27877) against the last commit this lane fully reviewed (232969a, job run) restricted to this PR's six changed files. server.py, test_server.py, README.md, schema/questions.schema.json, and plugin.json are byte-for-byte identical between the two commits; CHANGELOG.md only picked up one unrelated entry (the #5458 interview-wording-lint fix) from a main merge, not from this PR's own changes. The intervening commits on this branch (3c0a131, 232969a → 26e2787) are merges of origin/main; none touch plugins/planning/surface/server.py or the other files this PR changes.

Since the code this lane reviews is unchanged from the commit already reviewed with no blocking findings — including confirmation that Codex's P1 (reopen misreading answered: true) is fixed via the bool((latest_decision(...) or {}).get("decision")) check and covered by TestAnswered.test_4_reopen_is_not_answered — that review still applies. No new findings to post.

[claude]

Claude finished @kyle-sexton's task in 46s —— View job

---

Security review (/review:security-review)

• Gather context (PR diff, changed files, prior reviews)
• Check skip gate (closed PR / no security surface / already-reviewed head)
• Review changed files against criteria
• Post findings

Result: skip gate hit (criterion 3) — this head's diff content already has a successful security review that still applies.

gh pr diff 5483 at HEAD 26e27877eb2c38c85a0e2775f659c15e6527f11d is byte-identical, for every changed file, to the diff already reviewed with no findings at 232969a4a4a85d17f14407c1c44eb6f85e5079b0 and 3c0a1310eb47d050a635e0c59f750c48512e5c44 (both linked above in this thread). This head is a merge of origin/main into the branch (see 26e2787); the incoming main commits (ef7c33f, d8b2be8, b0f045b, 58150fc) touch instruction-placement, planning's wording lint, map-states retirement, and repo-hygiene preflight scoping — none of them under plugins/planning/surface/ or plugins/planning/.claude-plugin/, so the merge changed nothing in this PR's own diff.

Re-confirming the substance for the record, since the merge is new even though the diff isn't:

• server.py#L805-L820 — Hub.state() computes answered via bool((latest_decision(x, r.get("responses", {})) or {}).get("decision")) over questions.json/responses.json, both already loaded locally by this handler. No new endpoint, no new input parsing, no change to /api/state's auth posture, no change to the token-guarded POST /api/answer/POST /api/lease paths. This is the fixed form of Codex's earlier P1 (reopen previously read answered: true off record presence alone); a reopen event carries decision: None, so bool(...) on it is False.
• exporters.latest_decision/newest_decision are unchanged by this PR — pure selection over data the handler already had.
• README.md#L71, schema/questions.schema.json#L246-L253, CHANGELOG.md, and the plugin.json version bump are documentation/metadata only.
• test_server.py's TestAnswered (4 cases) is test-only, including the reopen regression case.
• No CLAUDE.md/AGENTS.md/rules/skill/agent/hook file is touched in this diff, so the instruction-surface-deletion lens doesn't apply here.
• No workflow files changed, so Actions/zizmor hardening is out of scope regardless.

No findings. Nothing posted as inline comments.

[claude]

Claude finished @kyle-sexton's task in 39s —— View job

---

Security review (/review:security-review)

• Gather context (PR diff, changed files, prior reviews)
• Invoke /review:security-review per the custom instructions
• Post findings

Result: skip gate hit (criterion 3) — this head's diff content already has a successful security review that still applies.

I diffed the current HEAD (df42453d4e8928253f71da11ad03c7cbad903d04) against the last commit this lane fully reviewed with no findings (26e27877, job run), restricted to this PR's six changed files. server.py, test_server.py, README.md, schema/questions.schema.json, and plugin.json are byte-for-byte identical between the two commits; CHANGELOG.md only picked up one unrelated entry (the #5484 carry-chip wording fix) from the new origin/main merge, not from this PR's own changes. df42453 is itself just a merge of origin/main into this branch (see the commit's parents: e17d807, 2993d22, 7708cba, 8ecaf60); none of those incoming commits touch plugins/planning/surface/ or plugins/planning/.claude-plugin/.

Re-confirming the substance for the record, since the merge commit is new even though the diff isn't:

• server.py#L805-L820 — Hub.state() computes answered via bool((latest_decision(x, r.get("responses", {})) or {}).get("decision")) over questions.json/responses.json, both already loaded locally by this handler. No new endpoint, no new input parsing, no change to /api/state's auth posture, no change to the token-guarded POST /api/answer/POST /api/lease paths. This is the fixed form of Codex's earlier P1 (reopen previously read answered: true off record presence alone); a reopen event carries decision: None, so bool(...) on it is False.
• exporters.latest_decision/newest_decision are unchanged by this PR — pure selection over data the handler already had.
• README.md#L71, schema/questions.schema.json#L246-L253, CHANGELOG.md, and the plugin.json version bump are documentation/metadata only.
• test_server.py's TestAnswered (4 cases) is test-only, including the reopen regression case.
• No CLAUDE.md/AGENTS.md/rules/skill/agent/hook file is touched in this diff, so the instruction-surface-deletion lens doesn't apply here.
• No workflow files changed, so Actions/zizmor hardening is out of scope regardless.

No findings. Nothing posted as inline comments.

[claude]

Claude finished @kyle-sexton's task in 44s —— View job

---

Code review (/review:code-review) — PR #5483

• Gather context (PR diff, changed files, prior review comments)
• Invoke /review:code-review and follow its instructions
• Post findings

Skip gate hit — criterion 4 (a successful review from this lane already applies to this head).

Diffed current HEAD (df42453) against the last commit this lane fully reviewed (232969a, byte-identical per the prior pass at 26e2787), restricted to this PR's six files. server.py, test_server.py, README.md, schema/questions.schema.json, and plugin.json are unchanged. CHANGELOG.md only picked up unrelated entries from main merges (26e2787 → df42453 brought in the #5484 carry-chip wording fix and other unrelated fixes); none of the intervening merge commits (e529f55, 8ecaf60, 7708cba, 2993d22, e17d807) touch plugins/planning/surface/ or this PR's other changed files.

Since the reviewed code is unchanged from the commit already covered with no blocking findings — including confirmation that Codex's P1 (reopen misreading answered: true) is fixed via bool((latest_decision(...) or {}).get("decision")) in server.py#L807-L816 and covered by TestAnswered.test_4_reopen_is_not_answered in test_server.py — that review still applies. No new findings, and nothing posted as inline comments.

[claude]

Claude finished @kyle-sexton's task in 1m 33s —— View job

---

Security review (/review:security-review)

• Gather context (PR diff, changed files, prior reviews)
• Invoke /review:security-review per the custom instructions
• Post findings

Result: skip gate hit (criterion 3) — this head's diff content already has a successful security review that still applies.

Current HEAD is 22074de1f9b0d3260b858c096a2b320e43d4f86c, one commit past the b2f08be this task's header names (b2f08be is itself Merge origin/main into fix/5459-api-state-answered; 22074de on top of it just deletes the leftover <<<<<<</|||||||/======= conflict markers in plugins/planning/.claude-plugin/plugin.json, resolving them to "version": "0.48.0").

I diffed this HEAD against the last commit this lane fully reviewed with no findings (df42453, job run), restricted to this PR's six changed files:

• server.py, test_server.py, README.md, and plugin.json are byte-for-byte identical between df42453 and current HEAD — including plugin.json's version field, which is 0.48.0 on both sides of the merge-marker cleanup, so that commit is a no-op for content, not a real version change.
• schema/questions.schema.json and CHANGELOG.md only picked up unrelated entries from origin/main merges (the summaryOf schema property and changelog entry for fix(planning): flag an interview group summary as stale when questions are added to the group #5451/fix(planning): flag a group summary written before questions were added #5490, a group-summary-staleness fix) — not from this PR's own commits.

Re-confirming the substance for the record, since the commit is new even though the reviewed content isn't:

• server.py#L805-L820 — Hub.state() computes answered via bool((latest_decision(x, r.get("responses", {})) or {}).get("decision")) over questions.json/responses.json, both already loaded locally by this handler. No new endpoint, no new input parsing, no change to /api/state's auth posture, no change to the token-guarded POST /api/answer/POST /api/lease paths. This is the fixed form of Codex's earlier P1 (reopen previously read answered: true off record presence alone); a reopen event carries decision: None, so bool(...) on it is False.
• exporters.latest_decision/newest_decision are unchanged by this PR — pure selection over data the handler already had.
• README.md#L71 and schema/questions.schema.json document answered accurately; CHANGELOG.md and the plugin.json version bump are documentation/metadata only.
• test_server.py's TestAnswered (4 cases) is test-only, including the reopen regression case.
• No CLAUDE.md/AGENTS.md/rules/skill/agent/hook file is touched anywhere in this branch's history for this PR, so the instruction-surface-deletion lens doesn't apply.
• No workflow files changed, so Actions/zizmor hardening is out of scope regardless.

No findings. Nothing posted as inline comments.

[claude]

Claude finished @kyle-sexton's task in 1m 14s —— View job

---

Code review (/review:code-review) — PR #5483

• Gather context (PR diff, changed files, prior review comments)
• Check skip gate against HEAD 22074de
• Review changed files
• Post findings

Skip gate hit — criterion 4, extended by one trivial, verified-correct commit.

The core change (server.py's answered flag, schema/README/changelog docs, test_server.py) is byte-for-byte identical to the commit this lane already reviewed with no blocking findings (df42453, job run) — including confirmation that Codex's P1 (reopen misreading answered: true) is fixed via bool((latest_decision(...) or {}).get("decision")) and covered by TestAnswered.test_4_reopen_is_not_answered.

The only change since that head is 22074de, which resolves leftover merge-conflict markers a main merge (b2f08be) left in plugins/planning/.claude-plugin/plugin.json:

-<<<<<<< HEAD
   "version": "0.48.0",
-||||||| e17d8076b
-  "version": "0.47.6",
-=======
-  "version": "0.47.7",
->>>>>>> origin/main

I checked this resolution directly rather than trusting the diff alone:

• plugin.json#L4 now reads a single clean "version": "0.48.0", matching the top CHANGELOG.md entry (CHANGELOG.md#L6), which is this PR's own answered-flag entry — the correct resolution (not the incoming main value from the unrelated #5451 group-summary-staleness fix, which the merge already folded in separately as its own changelog entry at 0.47.7).
• Grepped the full plugins/planning/ tree for <<<<<<</=======/>>>>>>> markers: none remain anywhere.

The unrelated round.py/index.html/test_round.py/fixture changes visible between the last-reviewed head and this one are pre-existing content from the origin/main merge (#5490, already reviewed and merged on main), not part of this PR's own six changed files, and out of scope here.

No findings.

[claude]

Claude finished @kyle-sexton's task in 1m 21s —— View job

---

Security review (/review:security-review)

• Gather context (PR diff, changed files, prior reviews)
• Invoke /review:security-review per the custom instructions
• Post findings

Result: skip gate hit (criterion 3) — this head's diff content already has a successful security review that still applies.

New HEAD is 22074de, which only resolves a merge-conflict in plugin.json left over from merging origin/main:

-<<<<<<< HEAD
   "version": "0.48.0",
-||||||| e17d8076b
-  "version": "0.47.6",
-=======
-  "version": "0.47.7",
->>>>>>> origin/main

That keeps this PR's own version bump (0.48.0) and drops the conflict markers — metadata only, no code or logic. I confirmed no stray <<<<<<</=======/>>>>>>> markers remain anywhere under plugins/planning/.

Diffing this PR's own files between the previously-reviewed head (df42453, job run) and current HEAD:

• server.py, test_server.py, and README.md are byte-for-byte identical — no changes at all.
• schema/questions.schema.json and CHANGELOG.md only picked up an unrelated summaryOf/group-summary-staleness entry from a main merge (#5451, v0.47.7) — not from this PR's own diff.
• plugin.json changed only in the conflict-marker resolution shown above.

Re-confirming the substance for the record, since HEAD is new even though the reviewed code isn't:

• server.py#L802-L818 — Hub.state() computes answered via bool((latest_decision(x, r.get("responses", {})) or {}).get("decision")) over questions.json/responses.json, both already loaded locally by this handler. No new endpoint, no new input parsing, no change to /api/state's auth posture, no change to the token-guarded POST /api/answer/POST /api/lease paths. This is the fixed form of Codex's earlier P1 (reopen previously read answered: true off record presence alone); a reopen event carries decision: None, so bool(...) on it is False.
• exporters.latest_decision/newest_decision are unchanged by this PR — pure selection over data the handler already had.
• README.md, schema/questions.schema.json, CHANGELOG.md, and the plugin.json version bump are documentation/metadata only.
• test_server.py's TestAnswered (4 cases) is test-only, including the reopen regression case.
• No CLAUDE.md/AGENTS.md/rules/skill/agent/hook file is touched in this diff, so the instruction-surface-deletion lens doesn't apply here.
• No workflow files changed, so Actions/zizmor hardening is out of scope regardless.

No findings. Nothing posted as inline comments.

[claude]

Claude finished @kyle-sexton's task in 1m 1s —— View job

---

Security review (/review:security-review)

• Gather context (PR diff, changed files, prior reviews)
• Check skip gate against current HEAD
• Review changed files against criteria
• Post findings

Result: skip gate hit (criterion 3) — this head's diff content already has a successful security review that still applies.

Current HEAD is 0872c65, a merge of origin/main into this branch. I diffed it against the last commit this lane fully reviewed with no findings (22074de, job run), restricted to this PR's six changed files:

• plugins/planning/surface/server.py, test_server.py, schema/questions.schema.json, and .claude-plugin/plugin.json are byte-for-byte identical between 22074de and 0872c65.
• plugins/planning/CHANGELOG.md and surface/README.md only picked up an unrelated entry/table row from the origin/main merge (#4653, the new surface/DEFERRED.md doc) — not from this PR's own commits.
• The incoming main commits folded in by this merge (8e525fe, dab0b0a, 0f2a19d, 408f9a2, 9bedd89, f75bebe, 7b867ac, e24b48a, 43d8c35, and others back to 22074de) touch disk-hygiene, guardrails, work-items docs, the wizard, repo-hygiene, claude-ops, planning's deferred-list doc, plugin-quality, and claude-config — none of them under plugins/planning/surface/ or plugins/planning/.claude-plugin/.
• No <<<<<<</=======/>>>>>>> conflict markers remain anywhere under plugins/planning/; plugin.json's version is a clean "0.48.0".

Re-confirming the substance for the record, since HEAD is new even though the reviewed content isn't:

• server.py#L802-L818 — Hub.state() computes answered via bool((latest_decision(x, r.get("responses", {})) or {}).get("decision")) over questions.json/responses.json, both already loaded locally by this handler. No new endpoint, no new input parsing, no change to /api/state's auth posture, no change to the token-guarded POST /api/answer/POST /api/lease paths. This is the fixed form of Codex's earlier P1 (reopen previously read answered: true off record presence alone); a reopen event carries decision: None, so bool(...) on it is False.
• exporters.latest_decision/newest_decision are unchanged by this PR — pure selection over data the handler already had.
• README.md#L71 and schema/questions.schema.json document answered accurately; CHANGELOG.md and the plugin.json version bump are documentation/metadata only.
• test_server.py's TestAnswered (4 cases) is test-only, including the reopen regression case.
• No CLAUDE.md/AGENTS.md/rules/skill/agent/hook file is touched anywhere in this branch's history for this PR, so the instruction-surface-deletion lens doesn't apply.
• No workflow files changed, so Actions/zizmor hardening is out of scope regardless.

No findings. Nothing posted as inline comments.

[claude]

Claude finished @kyle-sexton's task in 39s —— View job

---

Code review (/review:code-review) — PR #5483

• Gather context (PR diff, changed files, prior review comments)
• Invoke /review:code-review and follow its instructions
• Post findings

Skip gate hit — criterion 4 (a successful review from this lane already applies to this head).

Diffed current HEAD (0872c65) against the last commit this lane fully reviewed with no blocking findings (22074de, job run), restricted to this PR's six changed files:

• server.py, test_server.py, schema/questions.schema.json, and plugin.json are byte-for-byte identical between the two commits.
• CHANGELOG.md and README.md only picked up one unrelated entry each from the origin/main merge (surface/DEFERRED.md, #4653) — not from this PR's own changes.

Since the code this lane reviews is unchanged from the commit already covered with no blocking findings — including confirmation that Codex's P1 (reopen misreading answered: true) is fixed via bool((latest_decision(...) or {}).get("decision")) in server.py#L807-L816 and covered by TestAnswered.test_4_reopen_is_not_answered in test_server.py — that review still applies. No new findings, and nothing posted as inline comments.
· branch fix/5459-api-state-answered

[claude]

Claude finished @kyle-sexton's task in 54s —— View job

---

Security review (/review:security-review) — PR #5483

• Gather context (PR diff, changed files, prior reviews)
• Invoke /review:security-review per the custom instructions
• Post findings

Result: skip gate hit (criterion 3) — this head's diff content already has a successful security review that still applies.

Current HEAD is 2bd246a, one commit past 558a90f (Merge remote-tracking branch 'origin/main'). I diffed it against the last commit this lane fully reviewed with no findings (0872c65, job run), restricted to this PR's six changed files:

• plugins/planning/surface/server.py, test_server.py, README.md, and schema/questions.schema.json are byte-for-byte identical between 0872c65 and 2bd246a.
• plugins/planning/CHANGELOG.md only picked up one unrelated entry ([0.47.9], the plan-reviewer maxTurns fix, #4256) from the origin/main merge — not from this PR's own changes.
• plugins/planning/.claude-plugin/plugin.json changed only via 2bd246a itself, which drops a duplicate "version" key ("version": "0.47.9") that the merge left behind after "version": "0.48.0", keeping this PR's own version bump. Metadata-only, no code or logic — confirmed with git show 2bd246a (1 file, 1 deletion).

Re-confirming the substance for the record, since HEAD is new even though the reviewed code isn't:

• server.py#L802-L818 — Hub.state() computes answered via bool((latest_decision(x, r.get("responses", {})) or {}).get("decision")) over questions.json/responses.json, both already loaded locally by this handler. No new endpoint, no new input parsing, no change to /api/state's auth posture, no change to the token-guarded POST /api/answer/POST /api/lease paths. This is the fixed form of Codex's earlier P1 (reopen previously read answered: true off record presence alone); a reopen event carries decision: None, so bool(...) on it is False.
• exporters.latest_decision/newest_decision are unchanged by this PR — pure selection over data the handler already had.
• README.md#L71 and schema/questions.schema.json document answered accurately; CHANGELOG.md and the plugin.json version bump are documentation/metadata only.
• test_server.py's TestAnswered (4 cases) is test-only, including the reopen regression case.
• No CLAUDE.md/AGENTS.md/rules/skill/agent/hook file is touched anywhere in this branch's history for this PR, so the instruction-surface-deletion lens doesn't apply.
• No workflow files changed, so Actions/zizmor hardening is out of scope regardless.

No findings. Nothing posted as inline comments.
· branch fix/5459-api-state-answered

[claude]

Claude finished @kyle-sexton's task in 31s —— View job

---

Code review (/review:code-review) — PR #5483

• Gather context (PR diff, changed files, prior review comments)
• Check skip gate against HEAD 2bd246a
• Review changed files
• Post findings

Skip gate hit — criterion 4 (a successful review from this lane already applies to this head).

Diffed current HEAD (2bd246a) against the last commit this lane fully reviewed with no blocking findings (0872c65, job run), restricted to this PR's six changed files. Two commits sit on top of that head:

• 558a90f — merge of origin/main. Only plugins/planning/CHANGELOG.md picked up content, and it's an unrelated entry (0.47.9, the plan-reviewer maxTurns fix, #4256) — not from this PR's own changes. server.py, test_server.py, README.md, schema/questions.schema.json, and plugin.json are untouched by the merge.
• 2bd246a — drops a duplicate "version" key the merge left in plugin.json ("version": "0.48.0" followed by a stray "version": "0.47.9"). I verified the result directly: the file now parses as valid JSON with a single version field, "0.48.0", matching the top CHANGELOG.md entry (this PR's answered-flag entry). Correct fix, no logic involved.

Since the reviewed code (Hub.state()'s answered computation, its tests, and its docs) is byte-for-byte unchanged from the commit already covered with no blocking findings — including confirmation that Codex's earlier P1 (reopen misreading answered: true) is fixed via bool((latest_decision(...) or {}).get("decision")) in server.py#L807-L816 and covered by TestAnswered.test_4_reopen_is_not_answered in test_server.py — that review still applies. No new findings, and nothing posted as inline comments.
· branch fix/5459-api-state-answered

[claude]

Claude finished @kyle-sexton's task in 36s —— View job

---

Code review (/review:code-review) — PR #5483

• Gather context (PR diff, changed files, prior review comments)
• Check skip gate against HEAD 5093bdb
• Diff this PR's six files against the last commit this lane reviewed
• Post findings

Skip gate hit — criterion 4 (a successful review from this lane already applies to this head).

Diffed current HEAD (5093bdba7bb70fdf7df840dce57b04dfe8ab1cb3, Merge remote-tracking branch 'origin/main' into fix/5459-api-state-answered) against the last commit this lane fully reviewed with no blocking findings (2bd246a, job run), restricted to this PR's six changed files:

• server.py, test_server.py, README.md, schema/questions.schema.json, and plugin.json are byte-for-byte identical between the two commits.
• CHANGELOG.md only picked up one unrelated entry from the origin/main merge ([0.47.10], the interview-defenses.test.sh attended-only wording fix, #5332) — not from this PR's own changes.

Since the code this lane reviews — Hub.state()'s answered computation, its tests, and its docs — is unchanged from the commit already covered with no blocking findings, including confirmation that Codex's earlier P1 (reopen misreading answered: true) is fixed via bool((latest_decision(...) or {}).get("decision")) in server.py#L807-L816 and covered by TestAnswered.test_4_reopen_is_not_answered in test_server.py, that review still applies. No new findings, and nothing posted as inline comments.
· branch fix/5459-api-state-answered

[claude]

Claude finished @kyle-sexton's task in 55s —— View job

---

Security review (/review:security-review) — PR #5483

• Gather context (PR diff, changed files, prior reviews)
• Check skip gate against HEAD 5093bdb
• Review changed files against criteria
• Post findings

Result: skip gate hit (criterion 3) — this head's diff content already has a successful security review that still applies.

Current HEAD is 5093bdb, a merge of origin/main into this branch. I diffed it against the last commit this lane fully reviewed with no findings (2bd246a, job run), restricted to this PR's six changed files:

• plugins/planning/surface/server.py, test_server.py, README.md, schema/questions.schema.json, and .claude-plugin/plugin.json are byte-for-byte identical between 2bd246a and 5093bdb (git diff --stat empty).
• plugins/planning/CHANGELOG.md only picked up one unrelated entry ([0.47.10], the interview-defenses attended-only re-pin note, #5332) from the origin/main merge — not from this PR's own commits.
• The incoming main commits folded in by this merge (ab23b4a, 8b9ee50, 271684f, 3891145, a51dd18, 84548fe, c3baa96, f9929d0, 6fa0d73, ade6eff, f1f47db) touch claude-ops script auditing, docs/interop records, skill-description trimming, config-location docs, the interview-defenses test, disk-hygiene docs, guardrails, instruction-placement, discovery setup, and the security-reviewer's instruction-surface-deletion lens. Only 84548fe touches a file this PR also changes (plugin.json, bumping main's 0.47.9 → 0.47.10); the merge resolved cleanly, keeping this PR's own 0.48.0 (higher than either side), matching the top CHANGELOG.md entry — this PR's own answered-flag entry. Metadata-only, no logic.
• No CLAUDE.md/AGENTS.md/rules/skill/agent/hook file is touched by this PR's own diff, so the instruction-surface-deletion lens f1f47db adds doesn't flag anything here.
• No workflow files changed, so Actions/zizmor hardening is out of scope regardless.

Re-confirming the substance for the record, since HEAD is new even though the reviewed content isn't:

• server.py#L802-L818 — Hub.state() computes answered via bool((latest_decision(x, r.get("responses", {})) or {}).get("decision")) over questions.json/responses.json, both already loaded locally by this handler. No new endpoint, no new input parsing, no change to /api/state's auth posture, no change to the token-guarded POST /api/answer/POST /api/lease paths. This is the fixed form of Codex's earlier P1 (reopen previously read answered: true off record presence alone); a reopen event carries decision: None, so bool(...) on it is False.
• exporters.latest_decision/newest_decision are unchanged by this PR — pure selection over data the handler already had.
• README.md#L71 and schema/questions.schema.json document answered accurately; CHANGELOG.md and the plugin.json version bump are documentation/metadata only.
• test_server.py's TestAnswered (4 cases) is test-only, including the reopen regression case.

No findings. Nothing posted as inline comments.
· branch fix/5459-api-state-answered

[claude]

Claude finished @kyle-sexton's task in 52s —— View job

---

Code review (/review:code-review) — PR #5483

• Gather context (PR diff, changed files, prior review comments)
• Invoke /review:code-review and follow its instructions
• Post findings

Skip gate hit — criterion 4 (a successful review from this lane already applies to this head).

Current HEAD is 006e9ee (Merge remote-tracking branch 'origin/main' into fix/5459-api-state-answered). I diffed this PR's own changed files (server.py, test_server.py, README.md, schema/questions.schema.json, CHANGELOG.md, plugin.json) against the merge base with origin/main (git diff 0d523c68..HEAD) and confirmed it's exactly the 54/2 diff this lane already reviewed with no blocking findings at prior heads (most recently 5093bdb, job run):

• server.py#L805-L820 — Hub.state()'s answered computation, bool((latest_decision(x, r.get("responses", {})) or {}).get("decision")), is byte-for-byte unchanged. This is the fixed form of Codex's earlier P1 (reopen misreading answered: true): a reopen event carries decision: None, so bool(...) on it is False, and TestAnswered.test_4_reopen_is_not_answered in test_server.py covers it.
• schema/questions.schema.json#L254-L257 and README.md#L73 document answered accurately and unchanged from the prior review.
• CHANGELOG.md carries this PR's own [0.48.2] entry for the answered flag; the intervening [0.48.1], [0.47.12], [0.47.11] entries are unrelated features (wake.sh, the assumptions-count hint, interview Step 3 recap) folded in by the origin/main merges, not this PR's own changes.
• plugin.json's version (0.48.2) matches the top CHANGELOG.md entry; the bump from 0.48.0→0.48.2 across the merges is from those unrelated main entries, not a discrepancy in this PR.

No new findings, and nothing posted as inline comments.
· branch fix/5459-api-state-answered

[claude]

Claude finished @kyle-sexton's task in 43s —— View job

---

Security review (/review:security-review) — PR #5483

• Gather context (PR diff, changed files, prior reviews)
• Check skip gate against HEAD 006e9ee
• Review changed files against criteria
• Post findings

Result: skip gate hit (criterion 3) — this head's diff content already has a successful security review that still applies.

Current HEAD is 006e9ee, a merge of origin/main into this branch. I diffed it against the last commit this lane fully reviewed with no findings (5093bdb, job run), restricted to this PR's six changed files:

• plugins/planning/surface/server.py, test_server.py, and schema/questions.schema.json are byte-for-byte identical between 5093bdb and 006e9ee (git diff --stat empty).
• plugins/planning/.claude-plugin/plugin.json only changed its version field (0.48.0 → 0.48.2), following the plugin's own version bump convention for the intervening main merges — metadata only.
• plugins/planning/CHANGELOG.md and surface/README.md only picked up unrelated entries from the origin/main merge ([0.48.1]/[0.47.12]/[0.47.11] — the Wrap-up assumptions warning, the interview Step 3 sign-off recap, the watcher wake.sh wrapper, and the watcher-JSON docs fix) — none from this PR's own commits.
• No <<<<<<</=======/>>>>>>> conflict markers remain anywhere under plugins/planning/.

Re-confirming the substance for the record, since HEAD is new even though the reviewed content isn't:

• server.py#L802-L818 — Hub.state() computes answered via bool((latest_decision(x, r.get("responses", {})) or {}).get("decision")) over questions.json/responses.json, both already loaded locally by this handler. No new endpoint, no new input parsing, no change to /api/state's auth posture, no change to the token-guarded POST /api/answer/POST /api/lease paths. This is the fixed form of Codex's earlier P1 (reopen previously read answered: true off record presence alone); a reopen event carries decision: None, so bool(...) on it is False.
• exporters.latest_decision/newest_decision are unchanged by this PR — pure selection over data the handler already had.
• README.md#L71 and schema/questions.schema.json document answered accurately; CHANGELOG.md and the plugin.json version bump are documentation/metadata only.
• test_server.py's TestAnswered (4 cases) is test-only, including the reopen regression case.
• No CLAUDE.md/AGENTS.md/rules/skill/agent/hook file is touched anywhere in this branch's history for this PR, so the instruction-surface-deletion lens doesn't apply.
• No workflow files changed, so Actions/zizmor hardening is out of scope regardless.

No findings. Nothing posted as inline comments.
· branch fix/5459-api-state-answered

[claude]

Claude finished @kyle-sexton's task in 1m 10s —— View job

---

Security review (/review:security-review) — PR #5483

• Gather context (PR diff, changed files, prior reviews)
• Check skip gate against HEAD 35257d5
• Review changed files against criteria
• Post findings

Result: skip gate hit (criterion 3) — this head's diff content already has a successful security review that still applies.

Current HEAD is 35257d5, a merge of origin/main into this branch. This PR's own diff is git diff 5eb1b99...HEAD (merge-base with origin/main), scoped to the same six files as before, 54 insertions / 2 deletions — matching the PR description exactly:

• plugins/planning/surface/server.py, test_server.py, README.md, and schema/questions.schema.json are byte-for-byte identical to the content this lane already reviewed with no findings at prior heads (most recently 006e9ee, job run).
• plugins/planning/.claude-plugin/plugin.json only bumped its version (0.48.2 → 0.49.1) and CHANGELOG.md only re-topped this PR's own answered-flag entry above a newer, unrelated [0.49.0] entry (revise --commit, #5497) that landed on main and was folded in by the merge — not from this PR's own commits. Metadata/docs only.
• The intervening merge brought in the revise --commit feature (commitments, commitsSinceSeq, the confirm contentRev-staleness check in server.py around repeat_of/Hub.route), but that code is already merged and reviewed on main (commit 5eb1b99) — it is not part of this PR's own diff, so it's out of scope for this lane's review of PR fix(planning): expose per-question answered flag in /api/state #5483.

Re-confirming the substance for the record, since HEAD is new even though the reviewed content isn't:

• server.py#L805-L820 — Hub.state() computes answered via bool((latest_decision(x, r.get("responses", {})) or {}).get("decision")) over questions.json/responses.json, both already loaded locally by this handler. No new endpoint, no new input parsing, no change to /api/state's auth posture, no change to the token-guarded POST /api/answer/POST /api/lease paths. This is the fixed form of Codex's earlier P1 (reopen previously read answered: true off record presence alone); a reopen event carries decision: None, so bool(...) on it is False.
• exporters.latest_decision/newest_decision are unchanged by this PR — pure selection over data the handler already had.
• README.md#L73 and schema/questions.schema.json#L258-L261 document answered accurately; CHANGELOG.md and the plugin.json version bump are documentation/metadata only.
• test_server.py's TestAnswered (4 cases) is test-only, including the reopen regression case.
• No CLAUDE.md/AGENTS.md/rules/skill/agent/hook file is touched by this PR's own diff, so the instruction-surface-deletion lens doesn't apply.
• No workflow files changed, so Actions/zizmor hardening is out of scope regardless.

No findings. Nothing posted as inline comments.
· branch fix/5459-api-state-answered

[claude]

Claude finished @kyle-sexton's task in 45s —— View job

---

Code review (/review:code-review) — PR #5483

• Gather context (PR diff, changed files, prior review comments)
• Invoke /review:code-review and follow its instructions
• Post findings

Skip gate hit — criterion 4 (a successful review from this lane already applies to this head).

Current HEAD is 35257d5 (Merge remote-tracking branch 'origin/main' into fix/5459-api-state-answered). I compared this PR's own diff against origin/main (git diff origin/main...HEAD, scoped to the six files this PR changes) with the diff already reviewed with no blocking findings at the last several heads (most recently 006e9ee, job run):

• server.py#L807-L823 — Hub.state()'s answered computation, bool((latest_decision(x, r.get("responses", {})) or {}).get("decision")), is byte-for-byte unchanged. This is the fixed form of Codex's earlier P1 (reopen misreading answered: true): a reopen event carries decision: None, so bool(...) on it is False, and TestAnswered.test_4_reopen_is_not_answered in test_server.py covers it.
• schema/questions.schema.json#L258-L261 and README.md#L73 document answered accurately and unchanged from the prior review.
• CHANGELOG.md carries this PR's own [0.49.1] entry for the answered flag; the intervening [0.49.0] and earlier entries are unrelated features/fixes folded in by origin/main merges, not this PR's own changes.
• plugin.json's version (0.49.1) matches the top CHANGELOG.md entry; the bump path from 0.48.2 → 0.49.1 across merges is from those unrelated main entries, not a discrepancy in this PR.

test_server.py's TestAnswered (4 cases) is unchanged and still test-only, exercising unanswered, terminal-decision, page-accept, and the accept-then-reopen regression case.

No new findings, and nothing posted as inline comments.
· branch fix/5459-api-state-answered

[claude]

Claude finished @kyle-sexton's task in 44s —— View job

---

Code review (/review:code-review) — PR #5483

• Gather context (PR diff, changed files, prior review comments)
• Check skip gate against HEAD 5e8fad0
• Diff this PR's six files against the last commit this lane reviewed
• Post findings

Skip gate hit — criterion 4 (a successful review from this lane already applies to this head).

Diffed current HEAD (5e8fad0b56e76a837deba95b85178bad9577760a, Merge remote-tracking branch 'origin/main' into fix/5459-api-state-answered) against the last commit this lane fully reviewed with no blocking findings (35257d5, job run), restricted to this PR's six changed files:

• server.py, test_server.py, README.md, and schema/questions.schema.json are byte-for-byte identical between the two commits.
• CHANGELOG.md and plugin.json only picked up this repo's usual version-bump/re-top mechanics from the origin/main merge (0.49.1 → 0.49.2, folding in the unrelated #5461 export-brief-acceptance-criteria fix, e92880c) — not from this PR's own commits. This PR's own answered-flag entry is still the top-most changelog item under the new version header.
• No <<<<<<</=======/>>>>>>> conflict markers remain anywhere under plugins/planning/.

Since the code this lane reviews — Hub.state()'s answered computation, its tests, and its docs — is unchanged from the commit already covered with no blocking findings, including confirmation that Codex's earlier P1 (reopen misreading answered: true) is fixed via bool((latest_decision(...) or {}).get("decision")) in server.py#L807-L816 and covered by TestAnswered.test_4_reopen_is_not_answered in test_server.py, that review still applies. No new findings, and nothing posted as inline comments.
· branch fix/5459-api-state-answered

[claude]

Claude finished @kyle-sexton's task in 1m 0s —— View job

---

Security review (/review:security-review) — PR #5483

• Gather context (PR diff, changed files, prior reviews)
• Check skip gate against HEAD 5e8fad0
• Review changed files against criteria
• Post findings

Result: skip gate hit (criterion 3) — this head's diff content already has a successful security review that still applies.

Current HEAD is 5e8fad0, a merge of origin/main into this branch (parents 35257d5 and e92880c). I diffed it against the last commit this lane fully reviewed with no findings (35257d5, job run), restricted to this PR's six changed files:

• plugins/planning/surface/server.py, test_server.py, and README.md are byte-for-byte identical between 35257d5 and 5e8fad0.
• plugins/planning/surface/schema/questions.schema.json is unchanged.
• plugins/planning/.claude-plugin/plugin.json and CHANGELOG.md only picked up an unrelated version bump/entry from the origin/main merge ([0.49.2] — export-brief carrying restated acceptance criteria, #5461) — not from this PR's own commits.

I also checked whether the merge's changes elsewhere in plugins/planning/surface/ could affect the security-relevant behavior this PR introduces, since Hub.state()'s answered flag depends on exporters.latest_decision/newest_decision:

• The merge does change plugins/planning/surface/exporters.py (export_brief's acceptance-criteria rendering), but that change is confined to export_brief and does not touch latest_decision or newest_decision, the functions this PR's answered computation calls. Those two functions are byte-for-byte unchanged. test_exporters.py additions are test-only and cover export_brief, unrelated to answered.
• No <<<<<<</=======/>>>>>>> conflict markers remain anywhere under plugins/planning/.

The rest of this merge (.claude/hooks/hook-telemetry-sink.sh, docs/out-of-scope/* deletions/edits, plugins/claude-ops/**, plugins/testing/**) is pre-existing content already merged and reviewed on main in its own PRs — none of it is part of this PR's own six-file diff, so it's out of scope for this lane's review of PR #5483. (For the record: the docs/out-of-scope/*.md deletions in this merge are not this PR's own change, so the instruction-surface-deletion lens doesn't apply to this PR's diff.)

Re-confirming the substance for the record, since HEAD is new even though the reviewed content isn't:

• server.py#L807-L823 — Hub.state() computes answered via bool((latest_decision(x, r.get("responses", {})) or {}).get("decision")) over questions.json/responses.json, both already loaded locally by this handler. No new endpoint, no new input parsing, no change to /api/state's auth posture, no change to the token-guarded POST /api/answer/POST /api/lease paths. This is the fixed form of Codex's earlier P1 (reopen previously read answered: true off record presence alone); a reopen event carries decision: None, so bool(...) on it is False.
• README.md#L73 and schema/questions.schema.json#L258-L261 document answered accurately; CHANGELOG.md and the plugin.json version bump are documentation/metadata only.
• test_server.py's TestAnswered (4 cases) is test-only, including the reopen regression case.
• No CLAUDE.md/AGENTS.md/rules/skill/agent/hook file is touched by this PR's own diff, so the instruction-surface-deletion lens doesn't apply here.
• No workflow files changed by this PR's own diff, so Actions/zizmor hardening is out of scope regardless.

No findings. Nothing posted as inline comments.
· branch fix/5459-api-state-answered

[claude]

Claude finished @kyle-sexton's task in 50s —— View job

---

Security review (/review:security-review) — PR #5483

• Gather context (PR diff, changed files, prior reviews)
• Check skip gate against HEAD 2ebe645
• Review changed files against criteria
• Post findings

Result: skip gate hit (criterion 3) — this head's diff content already has a successful security review that still applies.

Current HEAD is 2ebe645, a merge of origin/main into this branch (the other new parent, 1142789, is #5568, which drops availability-assertion wording from two unrelated Boundary bullets in draft-goal-condition/plan — not under plugins/planning/surface/ or plugins/planning/.claude-plugin/). I diffed it against the last commit this lane fully reviewed with no findings (5e8fad0, job run), restricted to this PR's six changed files:

• plugins/planning/surface/server.py, test_server.py, README.md, and schema/questions.schema.json are byte-for-byte identical between 5e8fad0 and 2ebe645 (git diff --stat empty).
• plugins/planning/.claude-plugin/plugin.json and CHANGELOG.md only picked up this repo's usual version-bump/re-top mechanics from the merge (0.49.2 → 0.49.3), folding in the unrelated #5568 Boundary-bullet entry above this PR's own answered-flag entry, which stays the top-most substantive item.
• No <<<<<<</=======/>>>>>>> conflict markers remain anywhere under plugins/planning/.

#5568's own change (dropping "ships with Claude Code" assertions from two Boundary bullets) is not part of this PR's own diff, so it's out of scope for this lane's review of PR #5483; it also doesn't touch CLAUDE.md/AGENTS.md/a rules file/a hook's allow/deny list, so the instruction-surface-deletion lens wouldn't apply to it here regardless.

Re-confirming the substance for the record, since HEAD is new even though the reviewed content isn't:

• server.py#L807-L823 — Hub.state() computes answered via bool((latest_decision(x, r.get("responses", {})) or {}).get("decision")) over questions.json/responses.json, both already loaded locally by this handler. No new endpoint, no new input parsing, no change to /api/state's auth posture, no change to the token-guarded POST /api/answer/POST /api/lease paths. This is the fixed form of Codex's earlier P1 (reopen previously read answered: true off record presence alone); a reopen event carries decision: None, so bool(...) on it is False.
• exporters.latest_decision/newest_decision are unchanged by this PR — pure selection over data the handler already had.
• README.md#L73 and schema/questions.schema.json#L258-L261 document answered accurately; CHANGELOG.md and the plugin.json version bump are documentation/metadata only.
• test_server.py's TestAnswered (4 cases) is test-only, including the reopen regression case.
• No CLAUDE.md/AGENTS.md/rules/skill/agent/hook file is touched by this PR's own diff, so the instruction-surface-deletion lens doesn't apply here.
• No workflow files changed by this PR's own diff, so Actions/zizmor hardening is out of scope regardless.

No findings. Nothing posted as inline comments.
· branch fix/5459-api-state-answered

[claude]

Claude finished @kyle-sexton's task in 28s —— View job

---

Code review (/review:code-review) — PR #5483

• Gather context (PR diff, changed files, prior review comments)
• Check skip gate against HEAD 2ebe645
• Diff this PR's six files against the last commit this lane reviewed
• Post findings

Skip gate hit — criterion 4 (a successful review from this lane already applies to this head).

Diffed current HEAD (2ebe6452b0862f4ce358e4924ed92453c9b073b3, Merge remote-tracking branch 'origin/main' into fix/5459-api-state-answered) against the last commit this lane fully reviewed with no blocking findings (5e8fad0, job run), restricted to this PR's six changed files:

• server.py, test_server.py, README.md, and schema/questions.schema.json are byte-for-byte identical between the two commits (git diff --stat shows no changes to these four files).
• plugins/planning/.claude-plugin/plugin.json only bumped its version (0.49.2 → 0.49.3) and CHANGELOG.md only re-topped this PR's own answered-flag entry above a newer, unrelated [0.49.2] entry (the #5568 native-Boundary-bullet fix) folded in by the origin/main merge — not from this PR's own commits. This PR's own changelog entry (answered reporting per question) is still the top-most item under the new version header.

Since the code this lane reviews — Hub.state()'s answered computation, its tests, and its docs — is unchanged from the commit already covered with no blocking findings, including confirmation that Codex's earlier P1 (reopen misreading answered: true) is fixed via bool((latest_decision(...) or {}).get("decision")) in server.py#L807-L816 and covered by TestAnswered.test_4_reopen_is_not_answered in test_server.py, that review still applies. No new findings, and nothing posted as inline comments.
· branch fix/5459-api-state-answered

[claude]

Claude finished @kyle-sexton's task in 46s —— View job

---

Security review (/review:security-review) — PR #5483

• Gather context (PR diff, changed files, prior reviews)
• Check skip gate against HEAD 295e3c4
• Review changed files against criteria
• Post findings

Result: skip gate hit (criterion 3) — this head's diff content already has a successful security review that still applies.

Current HEAD is 295e3c4, a merge of origin/main into this branch (the other new parent, d1e8a11, is #5522, the interview-page title-badging feature — not under plugins/planning/surface/ or plugins/planning/.claude-plugin/). I diffed it against the last commit this lane fully reviewed with no findings (2ebe645, job run), restricted to this PR's six changed files:

• plugins/planning/surface/server.py, test_server.py, and schema/questions.schema.json are byte-for-byte identical between 2ebe645 and 295e3c4.
• plugins/planning/surface/README.md picked up one unrelated line from the origin/main merge (feat(planning): badge the interview page title while the tab is hidden #5522's title-badging note in the Watcher protocol paragraph) — not from this PR's own commits.
• plugins/planning/.claude-plugin/plugin.json and CHANGELOG.md only picked up this repo's usual version-bump/re-top mechanics from the merge (0.49.3 → 0.49.4), folding in feat(planning): badge the interview page title while the tab is hidden #5522's changelog entry above this PR's own answered-flag entry, which stays the top-most substantive item.
• This PR's own diff against the merge base with origin/main (d1e8a11) is exactly the 54 insertions / 2 deletions across the same six files described in the PR body — matching what this lane already reviewed.

#5522's own change (title badging while the tab is hidden) is not part of this PR's own diff, so it's out of scope for this lane's review of PR #5483; it also doesn't touch CLAUDE.md/AGENTS.md/a rules file/a hook's allow/deny list, so the instruction-surface-deletion lens wouldn't apply to it here regardless.

Re-confirming the substance for the record, since HEAD is new even though the reviewed content isn't:

• server.py#L807-L823 — Hub.state() computes answered via bool((latest_decision(x, r.get("responses", {})) or {}).get("decision")) over questions.json/responses.json, both already loaded locally by this handler. No new endpoint, no new input parsing, no change to /api/state's auth posture, no change to the token-guarded POST /api/answer/POST /api/lease paths. This is the fixed form of Codex's earlier P1 (reopen previously read answered: true off record presence alone); a reopen event carries decision: None, so bool(...) on it is False.
• exporters.latest_decision/newest_decision are unchanged by this PR — pure selection over data the handler already had.
• README.md#L73 and schema/questions.schema.json#L258-L261 document answered accurately; CHANGELOG.md and the plugin.json version bump are documentation/metadata only.
• test_server.py's TestAnswered (4 cases) is test-only, including the reopen regression case.
• No CLAUDE.md/AGENTS.md/rules/skill/agent/hook file is touched by this PR's own diff, so the instruction-surface-deletion lens doesn't apply here.
• No workflow files changed by this PR's own diff, so Actions/zizmor hardening is out of scope regardless.

No findings. Nothing posted as inline comments.
· branch fix/5459-api-state-answered

[claude]

Claude finished @kyle-sexton's task in 50s —— View job

---

Code review (/review:code-review) — PR #5483

• Gather context (PR diff, changed files, prior review comments)
• Check skip gate against HEAD 295e3c4
• Diff this PR's six files against the last commit this lane reviewed
• Post findings

Skip gate hit — criterion 4 (a successful review from this lane already applies to this head).

Current HEAD is 295e3c4 (Merge remote-tracking branch 'origin/main' into fix/5459-api-state-answered, merging in d1e8a11, the #5522 interview-title-badging feature). I diffed it against the last commit this lane fully reviewed with no blocking findings (2ebe645, job run), restricted to this PR's six changed files:

• plugins/planning/surface/server.py, test_server.py, and schema/questions.schema.json are byte-for-byte identical between the two commits (git diff --stat shows no changes to these three files).
• plugins/planning/.claude-plugin/plugin.json only bumped its version (0.49.3 → 0.49.4) and plugins/planning/CHANGELOG.md only re-topped this PR's own answered-flag entry above a newer, unrelated [0.49.3] entry (the #5522 interview-title-badging feature) folded in by the origin/main merge — not from this PR's own commits. This PR's own changelog entry (/api/state reports answered per question) is still the top-most item under the new version header.
• plugins/planning/surface/README.md only picked up one unrelated sentence from the same merge (the tab-title badging note in the watcher-protocol paragraph) — not from this PR's own changes.

Since the code this lane reviews — Hub.state()'s answered computation, its tests, and its schema/README docs — is unchanged from the commit already covered with no blocking findings, including confirmation that Codex's earlier P1 (reopen misreading answered: true) is fixed via bool((latest_decision(...) or {}).get("decision")) in server.py#L807-L816 and covered by TestAnswered.test_4_reopen_is_not_answered in test_server.py, that review still applies. No new findings, and nothing posted as inline comments.
· branch fix/5459-api-state-answered