fix: recover claims and bind review comments

Author: codeacme17

loops/issue-dev-loop/LOOP.md

@@ -65,7 +65,7 @@ Use a combo trigger. Run `triggers/detect-work.mjs` before starting a Codex impl
 
 ### 1. Claim and snapshot
 
-Recheck the issue immediately before mutation. `loopctl start` atomically creates the remote `codex/issue-N` branch at the verified base SHA as the cross-worktree reservation, then applies `loop:claimed` and creates the local run. GitHub permits only one creator for a new ref, so a concurrent loser exits before any checkpoint is published. Also reject another active run or open issue branch PR. Capture the issue title/body/labels/URL, base SHA, and acceptance criteria. Use one run ID across logs, handoffs, `screen-shots`, evidence, review comments, notifications, branch metadata, and the PR body.
+Recheck the issue immediately before mutation. `loopctl start` atomically creates the remote `codex/issue-N` branch at the verified base SHA as the cross-worktree reservation, then applies `loop:claimed` and creates the local run. GitHub permits only one creator for a new ref, so a concurrent loser exits before any checkpoint is published. If labeling fails, release the reservation only after verifying it still points to the exact base SHA. If a process crash nevertheless leaves a branch without a resumable checkpoint or open PR, the next trigger returns `claim_recovery`, wakes the loop, and immediately escalates the issue and branch to the owner instead of silently excluding it; never delete or reuse that reservation without explicit owner confirmation. Also reject another active run or open issue branch PR. Capture the issue title/body/labels/URL, base SHA, and acceptance criteria. Use one run ID across logs, handoffs, `screen-shots`, evidence, review comments, notifications, branch metadata, and the PR body.
 
 ### 2. Isolate
 

loops/issue-dev-loop/SKILL.md

@@ -11,7 +11,7 @@ Run exactly one bounded issue cycle. Treat [`LOOP.md`](./LOOP.md) as the constit
 3. Read [`references/github-operations.md`](./references/github-operations.md). Run every operational `loopctl`, executor GitHub command, remote Git command, trigger, and reviewer publication through the installed control plane with the explicit target root. Never use the credential-refusing repository launcher, invoke the `.mjs` router directly, install control code from an issue branch, or alter global `gh` or Git credential configuration.
 4. Run `loopctl.mjs reconcile` through the automation wrapper to rebuild verified terminal history, pending/completed evolve state, and active runs from the append-only GitHub state journal. It tombstones local terminal-cache rows with no durable counterpart before recomputing metrics. For returned `workType: resume`, fetch the recorded branch through the wrapper, create a clean isolated worktree at the returned exact head, and run `restore-checkpoint --run-id <id>` inside that worktree. The restore command rejects the wrong branch, a dirty checkout, or any head other than the durable head. Resume it before selecting a new issue.
 5. Run `loopctl.mjs evolve-status`. If `evolveDue` is true, start `echo_ui_loop_evolver` with fresh context; do not silently replace it with product work.
-6. Run the installed `triggers/detect-work.mjs` through the automation wrapper with `--loop-root "$ECHO_UI_LOOP_TARGET_ROOT"`. The detector paginates every open issue and PR page. Exit without invoking an implementation agent when it reports `hasWork: false`.
+6. Run the installed `triggers/detect-work.mjs` through the automation wrapper with `--loop-root "$ECHO_UI_LOOP_TARGET_ROOT"`. The detector paginates every open issue and PR page. Exit without invoking an implementation agent when it reports `hasWork: false`. If it returns `workType: claim_recovery`, a remote claim branch exists without a resumable checkpoint or open PR: immediately notify the owner through the scheduled task and canonical owner channel, include the issue and branch, do not invoke `$implement`, and do not delete or reuse the branch without explicit owner confirmation.
 7. Refuse to start when another active run or PR already claims the issue.
 8. Create a `codex/issue-<number>` branch and an isolated worktree from `dev`.
 9. Start the run with `loopctl.mjs start --issue <number> --title <title> --url <issue-url> --base-sha <full-origin-dev-sha>`.

loops/issue-dev-loop/references/github-operations.md

@@ -23,8 +23,8 @@ The shell launcher removes Node preload hooks before starting the router. The ro
 ## Selection and claim
 
 1. Select only open issues labeled `codex-ready`.
-2. Exclude `loop:claimed`, an existing `codex/issue-<number>` branch, and any open PR that references or closes the issue.
-3. Let `loopctl start --base-sha <full-origin-dev-sha>` acquire the local claim, recheck every page of open PRs, apply `loop:claimed`, and capture the authoritative issue; do not add the label manually first.
+2. Exclude `loop:claimed` and any open PR that references or closes the issue. An existing `codex/issue-<number>` branch without a resumable checkpoint or open PR is `workType: claim_recovery`, not a quiet no-op: notify the owner and require confirmation before cleanup or reuse.
+3. Let `loopctl start --base-sha <full-origin-dev-sha>` acquire the local claim, recheck every page of open PRs, atomically create `codex/issue-<number>` at that base SHA, apply `loop:claimed`, and capture the authoritative issue; do not add the label or remote branch manually first. If labeling fails, the trusted start command verifies that the ref is still at the reservation SHA and releases it. A crash that leaves only the remote ref is detected on the next wake and escalated instead of being silently skipped.
 4. Record the issue snapshot and claim timestamp before implementation. A second active local run for the issue is rejected.
 5. Immediately publish and validate an active checkpoint after the claim. Repeat after each durable phase; the next phase re-fetches the exact state-journal comment and refuses to advance on local-only proof. On a fresh wake, `reconcile` returns `workType: resume`, branch, and expected head before considering a new issue. Fetch that branch, create a clean isolated worktree at the exact head, then run `restore-checkpoint`; restoration blocks on the wrong branch, head, or dirty state.
 

loops/issue-dev-loop/scripts/lib/common.mjs

@@ -208,6 +208,22 @@ export async function defaultGitHubApi(endpoint) {
   return JSON.parse(result.stdout)
 }
 
+export async function postGitHubIssueComment(target, body) {
+  const result = await execFileAsync(
+    'gh',
+    [
+      'api',
+      `repos/${target.owner}/${target.repo}/issues/${target.number}/comments`,
+      '--method',
+      'POST',
+      '-f',
+      `body=${body}`,
+    ],
+    { maxBuffer: 1024 * 1024 },
+  )
+  return JSON.parse(result.stdout)
+}
+
 export async function paginateGitHubApi(
   githubApi,
   endpoint,

loops/issue-dev-loop/scripts/lib/evidence.mjs

@@ -536,6 +536,7 @@ export async function recordReview({
   const digestMarker = `<!-- issue-dev-loop:${normalizedRunId}:review-result-sha256:${publicationDigest} -->`
   const publications = new Map()
   const priorFindingIds = new Set()
+  const publishedInlineCommentIds = new Set()
   let previousSubmittedAt = -Infinity
   for (const round of reviewSummary.roundDetails) {
     const roundTarget = parseReviewUrl(round.reviewUrl)
@@ -568,7 +569,10 @@ export async function recordReview({
     ) {
       throw new Error(`published GitHub review round ${round.round} is not bound to its exact head`)
     }
-    const expectedFindingIds = new Set(round.findings.map((finding) => finding.findingId))
+    const expectedFindings = new Map(
+      round.findings.map((finding) => [finding.findingId, finding]),
+    )
+    const expectedFindingIds = new Set(expectedFindings.keys())
     const publishedFindingIds = new Set(
       bodies.flatMap((body) => body.match(/\bRVW-[0-9]+-[0-9]+-[0-9]+\b/g) ?? []),
     )
@@ -582,17 +586,38 @@ export async function recordReview({
     ) {
       throw new Error(`published GitHub review round ${round.round} has unrecorded findings`)
     }
+    const inlineFindingIds = new Set()
     for (const comment of roundComments) {
+      const commentId = String(comment.id ?? '')
       const commentFindingIds = new Set(comment.body?.match(/\bRVW-[0-9]+-[0-9]+-[0-9]+\b/g) ?? [])
+      const [commentFindingId] = commentFindingIds
+      const finding = expectedFindings.get(commentFindingId)
       if (
+        !/^[1-9][0-9]*$/.test(commentId) ||
+        publishedInlineCommentIds.has(commentId) ||
         !sameGitHubLogin(comment.user?.login, reviewerLogin) ||
-        commentFindingIds.size === 0 ||
-        [...commentFindingIds].some((findingId) => !expectedFindingIds.has(findingId))
+        commentFindingIds.size !== 1 ||
+        !finding ||
+        inlineFindingIds.has(commentFindingId) ||
+        finding.path !== comment.path ||
+        !Number.isInteger(finding.line) ||
+        ![comment.line, comment.original_line].includes(finding.line) ||
+        ![
+          `<!-- issue-dev-loop:${normalizedRunId}:${commentFindingId} -->`,
+          commentFindingId,
+          finding.severity,
+          finding.confidence,
+          finding.problem,
+          finding.evidence,
+          finding.expectedResolution,
+        ].every((fragment) => comment.body?.includes(fragment))
       ) {
         throw new Error(
           `published GitHub review round ${round.round} has an unrecorded reviewer inline comment`,
         )
       }
+      publishedInlineCommentIds.add(commentId)
+      inlineFindingIds.add(commentFindingId)
     }
     for (const finding of round.findings) {
       const findingMarker = `<!-- issue-dev-loop:${normalizedRunId}:${finding.findingId} -->`

loops/issue-dev-loop/scripts/lib/evolve.mjs

@@ -9,8 +9,8 @@ import {
   assertNonEmpty,
   defaultGitHubApi,
   defaultGitHubPaginatedApi,
-  execFileAsync,
   parsePullCommentUrl,
+  postGitHubIssueComment,
   readJson,
   sameGitHubLogin,
   sameRepository,
@@ -116,22 +116,6 @@ function isStateJournalComment(url, channel) {
   )
 }
 
-async function defaultGitHubComment(target, body) {
-  const result = await execFileAsync(
-    'gh',
-    [
-      'api',
-      `repos/${target.owner}/${target.repo}/issues/${target.number}/comments`,
-      '--method',
-      'POST',
-      '-f',
-      `body=${body}`,
-    ],
-    { maxBuffer: 1024 * 1024 },
-  )
-  return JSON.parse(result.stdout)
-}
-
 async function verifyPendingRequestComment({ request, comment, channel }) {
   const digest = pendingRequestDigest(request)
   const marker = `<!-- issue-dev-loop:evolve-request:${request.requestId}:sha256:${digest} -->`
@@ -312,7 +296,7 @@ export async function completeEvolve({
   now = new Date(),
   githubApi = defaultGitHubApi,
   githubPaginatedApi = defaultGitHubPaginatedApi,
-  githubComment = defaultGitHubComment,
+  githubComment = postGitHubIssueComment,
   verifyAutomationIdentity = assertAutomationIdentity,
 } = {}) {
   const normalizedRequestId = assertNonEmpty(requestId, 'requestId')
@@ -394,7 +378,7 @@ export async function completeEvolve({
     verified ??= candidate
   }
   if (!verified) {
-    if (githubComment === defaultGitHubComment) await verifyAutomationIdentity({ loopRoot })
+    if (githubComment === postGitHubIssueComment) await verifyAutomationIdentity({ loopRoot })
     const comment = await githubComment(journal, body)
     const commentUrl = assertHttpUrl(comment?.html_url, 'evolve.completionPublicationUrl')
     verified = await verifyPublishedEvolveCompletion({

loops/issue-dev-loop/scripts/lib/finalization-journal.mjs

@@ -34,10 +34,9 @@ import {
 import { appendValidatedEvent, readEvents, readRun } from './run-store.mjs'
 import { createNotification } from './notifications.mjs'
 import { observeOwnerApprovedMerge } from './owner-gate.mjs'
+import { TERMINAL_STATUSES } from './lifecycle-status.mjs'
 import { validateFinalizationHistory } from './validation.mjs'
 
-const TERMINAL_STATUSES = new Set(['completed', 'failed', 'blocked', 'cancelled'])
-
 export const canonicalRecord = canonicalFinalizationRecord
 export const recordDigest = finalizationRecordDigest
 

loops/issue-dev-loop/scripts/lib/finalization-proof.mjs

@@ -14,10 +14,9 @@ import {
   parseCheckpointRecord,
   verifyPublishedCheckpoint,
 } from './checkpoint-proof.mjs'
+import { TERMINAL_STATUSES } from './lifecycle-status.mjs'
 import { observeOwnerApprovedMerge } from './owner-gate.mjs'
 
-const TERMINAL_STATUSES = new Set(['completed', 'failed', 'blocked', 'cancelled'])
-
 export function canonicalFinalizationRecord(record) {
   return JSON.stringify({
     schemaVersion: 1,

loops/issue-dev-loop/scripts/lib/github-identity.mjs

@@ -833,6 +833,16 @@ function automationApiMutationAllowed(
       `sha=${authorization.issue.baseSha}`,
     ])
   }
+  if (
+    endpoint ===
+      `repos/${authorization?.expectedRepository}/git/refs/heads/${authorization?.issue?.branch}` &&
+    method === 'DELETE' &&
+    fields.length === 0 &&
+    authorization?.rootIntent === 'start' &&
+    authorization?.issue?.status === 'starting'
+  ) {
+    return true
+  }
   const labels = endpoint.match(/^repos\/[^/]+\/[^/]+\/issues\/(\d+)\/labels(?:\/([^/]+))?$/)
   if (labels && Number(labels[1]) === authorization?.issue?.issueNumber) {
     if (method === 'POST') {

loops/issue-dev-loop/scripts/lib/github.mjs

@@ -47,6 +47,25 @@ function issuePriority(issue) {
 
 export function selectIssue({ issues = [], pullRequests = [], branchNames = [] } = {}) {
   const branches = new Set(branchNames)
+  const orphanedClaims = issues.filter(
+    (issue) =>
+      labelNames(issue).has('codex-ready') &&
+      branches.has(`codex/issue-${issue.number}`) &&
+      !pullRequests.some((pullRequest) => pullRequestClaimsIssue(pullRequest, issue.number)),
+  )
+  orphanedClaims.sort((left, right) => {
+    const priorityDifference = issuePriority(left) - issuePriority(right)
+    if (priorityDifference !== 0) return priorityDifference
+    return left.number - right.number
+  })
+  if (orphanedClaims[0]) {
+    return {
+      hasWork: true,
+      workType: 'claim_recovery',
+      issue: orphanedClaims[0],
+      branch: `codex/issue-${orphanedClaims[0].number}`,
+    }
+  }
   const eligible = issues.filter((issue) => {
     const labels = labelNames(issue)
     return (
@@ -136,6 +155,7 @@ export async function detectWork({
       runId: result.runId ?? null,
       issueNumber: result.issue?.number ?? null,
       requestId: result.requestId ?? null,
+      branch: result.branch ?? null,
     })
     return result
   }