Daily Fro Bot Report — 2026-10-14 (UTC) #3992

Daily Fro Bot Report — 2026-10-14 (UTC)

Snapshot began 2026-10-14T04:27:19.707501+00:00. Categories 1–4 come from the separate remediation readback; this pass ran categories 5–8. No repair PR was produced.

Run Summary

✅ verified-clean · ⚠️/❌ warning/error · ❔ could-not-check. Unavailable, empty or denied evidence is never an all-clear.

Errored PRs

None. Remediation inspected #3991, #3989, #3988, #3987, and #3984: check runs and separate legacy commit statuses passed; all mergeable. Current rollups still show zero failed/pending checks and successful legacy statuses. No repair comments or repair PRs. The drafted-doc review consumer's semantic failure is below.

Security

Remediation evidence: three transitive development Dependabot alerts (two medium, one low), live npm audit three moderate/one low and zero high/critical, override-floor gate passes. @humanfs/node #59, fast-uri #77, KaTeX #79. All 17 external Action/reusable-workflow packages checked against GitHub Advisory Database; no affected pin. No security repair PR. Next: compatible Renovate updates; Scorecard's high aggregate rule label is not package-advisory severity.

Authority: npm registry audit advisories, GitHub Dependabot/Advisory Database, upstream tag-to-commit resolution. Cross-major affected ranges included; latest-major version drift was not surveyed by remediation.

Control-Plane Integrity

Remediation evidence: 31 workflows/two composites parse; 146 SHA/version references match 17 upstream tags; 74 production scripts load under Node 24 strip-only. Explicit App scopes, shared setup exceptions, same-repository data authority, privacy gates and live main/data protections checked. Surplus CodeQL reads remain. Execution guard is unavailable; root cause unresolved under #3979. No repair PR.

Code Quality

Remediation evidence: pnpm bootstrap, pnpm check-types, pnpm lint, pnpm test all pass: 102 files / 5,358 tests passed / three todo. This pass passed pnpm lint, direct pnpm exec tsc --noEmit --project ./tsconfig.json, and pnpm exec vitest run with the same counts. Install/build-writing wrappers were not repeated under working-dir/non-mutating-shell constraints.

Read-only wiki verification: all 47 content pages pass ingest wikilinks; no deterministic findings on this pass's edited page/index. Two existing broken markdown links remain on the unrelated repo page, plus four advisory findings. The first probe incorrectly included scaffold README files in the content-only ingest validator; corrected scope passed. git diff --check passes. Existing checkout changes were preserved.

Oversight

Paginated authenticated /user/repos and /user/orgs, then /orgs/{org}/repos for every returned organization: bfra-me, psware-ps2, pro-actions. Listing calls succeeded; psware-ps2 is empty, so provides no health evidence. “Can see” means returned repositories with at least read/pull permission. Only confirmed-public identities/findings are reported here.

Public snapshot: 222 open issues, 135 open PRs, 13 new issues in 24h (including the retiring daily report), 82 issues inactive >30d, 99 PRs aged >7d, 57 PRs inactive >14d, six unassigned bugs. Counts precede report creation/retirement. PR age uses creation; inactivity uses latest update, commit, review, issue/review comment, or state-change activity, read with pagination. No activity-source errors.

Top three hotspots, ranked by distinct qualifying finding URLs in this snapshot (overlaps count once; alert counts are not unique exploit root causes):

• marcusrbrown/extend-vscode: 99 — finding. Next: prioritize existing security work, then reconfirm inactive issue/PR queues.

• marcusrbrown/containers: 64 — finding. Next: prioritize existing security work, then reconfirm inactive issue/PR queues.

• marcusrbrown/sparkle: 64 — finding. Next: prioritize existing security work, then reconfirm inactive issue/PR queues.

New issues — next: triage each; the previous report is retired below: bfra-me/renovate-action/issues/3934, bfra-me/works/issues/5128, fro-bot/.github/issues/3990, fro-bot/agent/issues/1827, fro-bot/dashboard/issues/614, fro-bot/space-bus/issues/210, marcusrbrown/.dotfiles/issues/2876, marcusrbrown/infra/issues/1530, marcusrbrown/marcusrbrown.github.io/issues/466, marcusrbrown/mothership/issues/146, marcusrbrown/panthea/issues/225, marcusrbrown/systematic/issues/1113, marcusrbrown/tokentoilet/issues/1614

Unassigned bugs — next: reproduce and assign: bfra-me/ha-addon-repository/issues/569, marcusrbrown/infra/issues/1521, marcusrbrown/marcusrbrown.com/issues/517, marcusrbrown/marcusrbrown.com/issues/465, marcusrbrown/systematic/issues/1099, marcusrbrown/systematic/issues/740

Failing current default-head contexts — next: inspect failed steps/current inputs before retry: bfra-me/github-action/actions/runs/38468696657/job/115541879092, marcusrbrown/extend-vscode/actions/runs/38431679495/job/115415225242, marcusrbrown/marcusrbrown.github.io/actions/runs/38212153327/job/114689530095, marcusrbrown/marcusrbrown.github.io/actions/runs/38212153327/job/114688511911, marcusrbrown/marcusrbrown.github.io/actions/runs/38212153327/job/114688511657, marcusrbrown/panthea/actions/runs/38482805288/job/115589193055, marcusrbrown/vbs/actions/runs/38507506341/job/115668266384

Latest check per app/name and latest legacy status per context were inspected separately. These failures are current-head evidence, not freshly triggered tests. Empty check-run and legacy histories — next: confirm intended CI/archive coverage: fro-bot/tokentoilet, marcusrbrown/encode-2024-q3-grp14-jobseeker-ai, marcusrbrown/gala-chain-code, pro-actions/peter-murray_workflow-application-token-action.

Accessible public alerts: 253 Dependabot / 189 Code Scanning / zero secret alerts. Unavailable endpoints: 6/40, 25/40, 33/40 respectively; names/statuses are in the table. Next: confirm scanner configuration and token scopes without blanket credential expansion.

Gateway rollout awareness — category 8

#3512 remains Open; Project 1 is In Progress / waiting, consistent with incomplete acceptance. All 29 Project items were read. Six mismatches persist:

Dashboard #585 and #586: body Open / Project Todo, consistent. Remaining acceptance boxes are unchecked; completed source work is not authenticated production acceptance.

• Body live contract 1.8.0, fresh health 1.9.0.
• Body gateway v0.118.2, infra pin v0.119.2. Latest inspected router run skips gateway deploy: no newer deployment claim from its green result.
• Body dashboard 2026.10.15, infra source ledger 2026.10.21, successful actual deploy job 2026.10.24, completed October 13 23:11:26 UTC. Post-convergence logs identify active compose tag 2026.10.24 and a healthy running dashboard at digest sha256:efd71a845ba5bd419e5c31fbea3ec65353a873d2342a8687d5d20c96f10cdaba.
• Latest-release authority: GitHub Releases agent v0.122.1 / dashboard 2026.10.24. Tagged contract files declare 1.11.0 / 1.9.0. Minor contract/release drift included; no separate cross-major gateway survey.
• Latest tracker comment says v0.122.1 pending release; it published later on October 13 at 21:09:55 UTC. Its paired-adoption requirement remains valid. Consumer summary parser still rejects both new waiting statuses and skips invalid entries: bumping only its contract constant would silently drop waiting runs.

Next: dedicated tracker reconciles stale body/comment/ledger claims; dashboard adopts the full 1.11.0 wire vocabulary before a fixed compatible paired cutover. Authenticated SSE, question/approval settlement, checkout projection, cancellation, logout/CSRF and privacy acceptance were not checked. No tracker comment or Project edit.

Cross-Project Intelligence

Every metadata/repos.yaml row was considered, and every accessible entry's tree/workflows plus available manifest/agent guidance were inspected. Confirmed-public coverage: 32 trees / 224 workflow files; no truncated public tree or fetch error. Partial coverage: fro-bot/fro-bot.github.io and fro-bot/systematic have empty default workflow trees. Source inspection does not establish successful execution. Adoptable patterns:

1. Agent integration workflow: explicit working-dir/response-none delivery, trusted finalizer, single-job OIDC authority. Adopt those boundaries for future caller-owned delivery.
2. Systematic diagnostics: !cancelled() plus actual suite outcome keeps diagnostics alive after failure without turning failure into success. Adopt for producer/handoff diagnosis.
3. Infra alert validation: owner-only synthetic validation, paginated marker/title/label identity, duplicate fail-closed and exact bounded readback. Adopt validation-path checks; the ordinary production lookup remains bounded, so do not infer complete production dedup.

No implementation changes in this category.

Progressive Improvement

Authority: independently queried npm registry stable tags; major drift included. Manifest:

All 31 committed workflows are active. Current-head Main passes lint/types/tests, workflow/strip-only/dist checks; PR-only authority/mutation jobs correctly skip on push. Missing preset-validation gate remains #3793. No stale production TODO/FIXME in scripts/workflows: only a negative-test fixture and the oversight prompt's search instruction. Shared setup/pnpm conventions match documented exceptions. Prompt/config/guard uncertainty remains #3979.

Learning proposals: one open, zero >14d: #3985, approximately 2.1 days old. Eligible fro-bot[bot]-authored PR #3989, head docs/drafted-solutions, approximately 1.6 days / 38 hours old, explicitly Closes #3985. Zero uncovered proposals: neither stall rule fires. Harvest/draft/publish succeeded and the PR extends the expected solution doc. Human merge remains pending. Metrics #3674 is not authorship evidence.

Degraded review consumer: #3989 still has zero reviews/comments and requests fro-bot and marcusrbrown. Fresh log readbacks for first and second green jobs both show unauthorized_author: association CONTRIBUTOR is rejected before model review. Workflow admission and pinned v0.119.2 action authorization disagree. Do not call publication a successful automated review or this covered proposal an unauthored stall.

Needs Human Attention

1. Drafted-doc review admission: reuse #3989 and the rejected-job links. Paths: .github/workflows/fro-bot.yaml, .github/workflows/draft-solutions.yaml, scripts/drafted-solutions-publish.ts; upstream agent src/features/triggers/router.ts, skip-conditions-pr.ts, author-utils.ts, paired tests. Smallest safe fix: reconcile the exact same-repository App/drafted-head/verified-review-request trust path, or have an authorized human review. Do not admit all CONTRIBUTORS/bots, bypass fork/privacy gates, or repeatedly request the same rejected review. Verify spoof/fork/untrusted-sender negatives and a real review on the current head. Prior learning: docs/solutions/best-practices/agent-drafts-trusted-publish-split-2026-10-10.md.
2. Paired contract adoption and tracker reconciliation: #3512 remains the dedicated tracker. Paths: agent packages/gateway/src/operator-contract/version.ts; dashboard src/gateway/operator-contract/{version,run-summary}.ts, run-list/card consumers and notification service worker; infra apps/gateway/upstream.json, apps/dashboard/docker-compose.yaml, docs/runbooks/gateway-operator-auth-lifecycle.md. Adopt both waiting statuses, 1.10.0 additions, strict /operator#<runId> route grammar and fixtures before selecting fixed compatible releases. Reconcile six Open/Done mismatches, release availability, and body/source/deploy versions. Do not deploy the 1.11.0 producer against the 1.9.0 consumer or weaken exact matching. Verify summary retention, reconnect/settlement/terminal behavior and authenticated acceptance; preserve volume-aware paired rollback.
3. Remediation handoff — guard and CodeQL scopes: reuse #3979. Inspect effective redacted CI OpenCode initialization/configuration and .github/workflows/fro-bot.yaml; unavailable guard root cause remains unproven. Do not retry unavailable guard start/complete calls or widen credentials. Separately confirm pinned CodeQL requirements before removing surplus packages: read / actions: read in .github/workflows/codeql-analysis.yaml:27–31; preserve scan coverage and validate commands/YAML/action refs plus a CodeQL run. Prior learning: docs/solutions/workflow-issues/required-github-token-for-agent-steps-2026-06-22.md distinguishes required token wiring from token scope.
4. Fleet/CI coverage: each scanner HTTP error and empty check/workflow source is a detection limit. Verify configured coverage and least-privilege access on linked repositories. For the seven failing head contexts inspect logs and current inputs before retry; reuse existing root-cause issues/security PRs. Routine upgrades stay Renovate-owned. Reuse #3793 for a preset-validation gate through shared setup; preserve existing guards.
5. Durable knowledge handoff: this pass adds only to knowledge/wiki/repos/fro-bot--dashboard.md, knowledge/index.md, and append-only knowledge/log.md: closed-set consumer adoption and dated release/deploy/ledger boundaries. Prior overlay/rename changes and staged metadata are preserved; no metadata edit by this pass. Leave the wiki dirty for trusted caller ingestion; persistence is not yet verified. The sole response comment retires the previous daily report; no per-finding/tracker comments or Project mutations.

Run 38510579359 · oversight working-dir · caller owns trusted wiki ingestion.