Daily Fro Bot Report — 2026-10-05 (UTC) #3952

Daily Fro Bot Report — 2026-10-05 (UTC)

Run Summary

Status legend: ✅ verified-clean; ⚠️ warning; ❌ error; ❔ incomplete or unavailable source. No incomplete source is rendered clean.

Categories 1–4 are carried forward from the separate remediation comment, with current open-PR check readback. This oversight invocation made no remediation commits, branch changes, tracker writes, or fleet issue/PR edits.

Errored PRs

• Data promotion #3951, mise #3948, shared workflows #3946, undici #3941, and brace-expansion #3942: Lint remains failed. The two routine-update PRs additionally have failing legacy renovate/artifacts statuses. Next: resolve the existing security-integration dependency, then let Renovate refresh artifacts; do not rerun unchanged branches or clear statuses manually.
• Root cause, branch-execution exclusions, and remaining risks: remediation evidence. No duplicate security PR was created.

Security

• The separate pass checked available GitHub alerts and npm audit: 10 repository alerts, five high npm advisories, zero critical npm advisories. Existing fixes: undici 8.10.2 and brace-expansion 5.0.12. Each isolated PR remains blocked by the other package's global gate.
• Sources and scope: GitHub Advisory Database, npm registry verification in remediation evidence. Same-major patch remediation; major drift was not surveyed in that pass. Open fixes are not merged-default-branch remediation.

Control-Plane Integrity

• Remediation evidence: 29 workflows, two composites, 128 external references SHA-pinned with version comments; production scripts loaded under Node strip-only; live main protection retained strict checks/admin enforcement and denied force pushes/deletions.
• Authority decision has a narrower branch restriction than the documented data-only writer contract; Quartz setup diverges from the shared pnpm setup contract. Next steps are recorded in the remediation comment. Preserve the cache-isolated privacy job; its setup exception is intentional.

Code Quality

• Remediation validation: pnpm bootstrap, pnpm check-types, and pnpm test passed (88 files, 4,079 tests, three todo). pnpm lint failed on existing undici/brace-expansion floors; independent ESLint, Markdown links, solution examples, and committed-dist checks passed.
• The oversight wiki addition passed direct no-emit TypeScript, direct Vitest (same 4,079 passing tests), ESLint, Markdown-link/solution-example checks, and whitespace validation using cached dependencies. Install/build-writing wrappers were not repeated under this invocation's non-mutating-shell contract. No assertions, thresholds, or rules were weakened.

Oversight

Enumeration used paginated authenticated-account and organization repository listings (bfra-me, psware-ps2, pro-actions); all listing calls succeeded. Read access defines coverage. Public reporting contains only repositories whose live visibility was checked; no non-public identities or findings are reproduced.

Public snapshot before report rotation: 40 repositories; 234 open issues; 119 open PRs; 11 new open issues in the last 24 hours; 80 issues inactive over 30 days; five unassigned bugs; 83 PRs created over seven days ago; 52 PRs inactive over 14 days; 285 available Dependabot alerts; 201 Code Scanning security findings; seven failing latest check contexts on default-branch HEADs. Code Scanning returned 214 records; 13 informational/quality records without security severity are excluded from hotspot counts. Counts describe observed findings, not confirmed exploitability or independent root causes. PR inactivity uses paginated commit/review/comment/state-change evidence, rather than an issue-list updated_at value. Check pagination is complete and reruns are reduced to the latest check of each name/app; legacy statuses were queried separately.

Top three hotspots, ranked by the number of qualifying findings in this snapshot (one per qualifying issue, open PR, available alert, or failing latest check):

This ranks available qualifying records only; unavailable sources can change the true ordering. Sparkle has 50 observed findings including 15 high Dependabot alerts; agent also has 34 Code Scanning security findings despite zero Dependabot alerts. These channels are complementary, not interchangeable.

New-issue triage: renovate-action #3878, works #4981, dashboard #558, space-bus #200, dotfiles #2781, portfolio #453 / #451, mothership #136, panthea #123, tokentoilet #1584, plus the previous daily report being superseded. Next: triage each against existing work rather than opening duplicates.

Unassigned bugs: add-on #569, brand site #517 / #465, Systematic #1005 / #740. Next: establish reproduction and ownership for each; no assignments were changed by this pass.

❔ Security-alert endpoints returned HTTP 403 for fro-bot/tokentoilet, copiloting, cortexkit, jobseeker, gala-chain-code, and workflow-token action. Alert totals exclude these sources. Next: obtain read visibility or confirm scanning availability; do not infer zero alerts.

❔ Code Scanning returned HTTP 403 for five repositories and HTTP 404 for 20. Each affected repository and response is linked in the inventory below; 404 does not distinguish disabled scanning from unavailable access. Next: establish scanner/access availability, and review the existing linked findings rather than treating a missing endpoint as zero. Alert records include posture warnings and repeated image contexts; do not interpret them as distinct confirmed vulnerabilities.

❔ Default-branch check/status sources are empty for fro-bot/tokentoilet, jobseeker, gala-chain-code, and workflow-token action. Evidence is over 30 days old for domain holder, copiloting, and cortexkit. Next: verify intended activity/archival policy and CI liveness. An empty or old green source is not verified-clean.

Gateway rollout awareness: Project 1 currently has 20 items Done and #3512 In Progress, consistent with an unfinished rollout. Its body is stale: it calls dashboard #179 Open (actual CLOSED, July 11), still calls the revocation runbook #711 missing (actual CLOSED, July 12), and describes a gateway pin of v0.83.0 (current infra source: v0.113.2). The latest tracker comment says the dashboard deploy is waiting; that run is now completed/cancelled (September 29). Latest upstream agent release and operator source are v0.117.2 / contract 1.8.0, while the dashboard contract remains 1.6.0. These are source-pin and narrative mismatches, not proof of a live deployment mismatch. ❔ Live browser-session acceptance and current live contract were not verified. Tracker writes remain owned by the dedicated workflow.

Cross-Project Intelligence

Every entry in tracked metadata was considered, without filtering on onboarding status. 35/35 entries received a bounded tree/workflow/prompt scan (up to six relevant files initially, supplemented where skill files crowded out workflows). Name-based failures were recovered by stable GitHub node ID before declaring entries inaccessible. Inaccessible/unscanned entries after recovery: None. This is source sampling, not execution verification or a full repository audit. Public links only. Empty workflow sources: domain holder and generated Systematic site; do not infer a missing runtime daemon from these Pages-only layouts.

Adoptable findings:

• Infra's content/storage job separation: attacker-reachable content jobs have read permissions; schedule/main-only storage jobs hold OIDC/environment access and blocked-egress controls. Apply this boundary when extending control-plane jobs; keep privilege tied to job topology, not prompt assurances.
• Agent OSV workflow: separate introduced-vulnerability PR evidence from scheduled/full-tree reporting. Consider an additional posture channel here; do not replace or weaken the existing security floor gate.
• Release Alert: a trusted workflow_run observer verifies the failed Release job rather than treating any Main failure as publication failure. Adapt the observer/terminal-delivery principle to learning capture. Its label-filtered discovery is not a deduplication template to copy blindly.

Progressive Improvement

• Compounding stalled: ten open proposals, all over 14 days old at this snapshot. September 21 cohort (about 14 days): #3905, #3906, #3907, #3908, #3909. September 14 cohort (about 21 days): #3887, #3888, #3889, #3890, #3891. Expected next action for each is validated authorship into docs/solutions/, not another proposal or report.
• Delivery degraded: October 5 learning-capture run again has successful harvest/draft but missing capture-learnings-bodies and an ENOENT publisher failure, following September 28. Producer cause remains unproven. #3674 is newly ambiguous with pending backlog zero; that recurrence population does not include every unauthored proposal.
• Tool drift: authoritative npm latest comparisons include major drift: ESLint 10.11.0 → 10.12.0 (one minor, below the “more than a minor” trigger), Prettier 3.9.1 → 3.9.9 (patch), TypeScript 6.0.3 → 7.0.2 (major), Vitest 4.1.11 → 5.0.3 (major). The TypeScript compatibility hold is intentional, not an unexplained drift defect. Leave upgrades to Renovate and validate major compatibility after security integration; no version changes were made.
• Convention/annotations: the authority and setup findings from remediation remain open decisions. No actionable production TODO/FIXME was found; the sole script match is a test-diff fixture, and wiki historical annotations were excluded from code debt.
• Durable recurrence evidence was added to GitHub Actions CI knowledge, with index/log notes left dirty for caller-owned ingestion. Persistence is not claimed until that later step succeeds.

Needs Human Attention

1. Security integration: existing PRs/evidence identify pnpm-workspace.yaml / pnpm-lock.yaml and the whole-tree floor gate. Authorize a delivery strategy preserving dedicated review scope, verify the integrated tree and audit, then refresh Renovate artifacts. Do not suppress advisories, force-push, bypass protections, or retry unchanged branches.
2. Learning handoff and codification: failed run, workflow, and publisher. Diagnose the missing producer output; require parseable output before upload and verify privacy-gated deterministic publication. Distinguish intentional zero candidates from missing output. Preserve separate token scope. Separately author each of the ten linked proposals into docs/solutions/ and verify documentation gates; a transport fix alone does not clear this queue.
3. Rollout reconciliation: #3512 and Project 1. The body names closed work as open and old pins/deploy states as current. The dedicated tracker should reconcile these claims against the linked issue states and cancelled deploy; keep rollout In Progress until live acceptance evidence exists. Before changing apps/gateway/upstream.json, inspect apps/gateway/src/deploy.ts (currently no GATEWAY_OPERATOR_TRUSTED_PROXIES handling) and dashboard src/gateway/operator-contract/version.ts / stream consumers. Coordinate startup configuration and contract adoption with deployment; an Action-pin update does not update the daemon. Do not relax exact-match/privacy gates from this report path.
4. Coverage limits: six Dependabot endpoints deny read access; 25 Code Scanning endpoints return 403/404; four CI sources are empty and three carry only stale evidence. Restore or explicitly establish intended scan/CI availability before certifying these repositories clean. No settings, access grants, or environments were changed here.