Agentic OS Status
A live view of the Free For Charity Agentic OS: the automation backlog that sandboxed AI agents pick up, the pull requests in flight, deployment gates awaiting a human, and the latest entries from the Conductor log.
Updated yesterday () · FreeForCharity/FFC-Cloudflare-Automation · raw JSON
Gates awaiting approval
These workflow runs are paused at an environment approval gate. Only a human reviewer can approve them — agents never can.
Pull requests in flight
What counts: Open pull requests that are agent work on this backlog, across every repository listed in `repos` — the same org-wide scope as the backlog above, and as the pickup query in AGENTS.md. A PR is listed when it carries the agentic-os label, or when its body references an agentic-os issue in its OWN repository (e.g. 'Closes #123' / 'Refs #123'); a bare #123 is repo-local, so the same number in two repos is two different issues. Referenced-issue labels are what decide it — nothing labels the pull requests themselves. `open_prs_total` is the unfiltered open-PR count over exactly these same repositories.
Automation backlog
| Issue | Title | Assignee | Updated |
|---|---|---|---|
| #1193 | An agent-ready issue stays agent-ready after its defect is fixed: #1077 sat 6 days, and a file-changed sweep is too noisy to catch ittestingagent-ready | unassigned | yesterday |
| #719 | Agentic OS Conductor Log | unassigned | yesterday |
| #1080 | Security: freeze and guard free-text dispatch inputs interpolated into run: bodies (34 workflows, 20 gated on write)securityagent-ready | unassigned | yesterday |
| #1190 | fix(check-environment-protection): retry transport failures, never retry an HTTP answer (L244)agent-ready | unassigned | yesterday |
| #1185 | 739: a PR with unresolved review findings is invisible next to a PR waiting on a humanagent-ready | unassigned | yesterday |
| #1123 | Sweep for PRs that are green on a stale base and would fail the merge queueagent-ready | unassigned | yesterday |
| #1089 | Step 4's backlog band has been unmet for 10 consecutive runs and its only lever is disconnected: the queue is over-full with fresh work, not stale work | unassigned | yesterday |
| #949 | 🚨 Scheduled workflow failing: 228. WHMCS - Fraud Review (FraudLabs Pro) [FRAUDLABS+WHMCS]bug | unassigned | yesterday |
| #1182 | A tool-gated `sys.exit(0)` makes a test module report green having asserted nothingagent-ready | unassigned | yesterday |
| #1177 | docs: write down the data-delivery contract — the Conductor's status deliveries auto-merge, and the window they needagent-ready | unassigned | yesterday |
| #877 | Key Vault credential health: dead, expiring, or unverifiable credentialssecuritypriority: high | unassigned | yesterday |
| #921 | 🚨 Scheduled workflow failing: 502. Google - Analytics Report (GA4 -> JSON) [GOOGLE]bug | unassigned | yesterday |
| #1175 | 226 reconcile-report groups duplicates by `clientid` alone, so the same organization applying under two client accounts is structurally invisiblebugagent-ready | unassigned | yesterday |
| #1174 | pwsh step-body tests run bodies as `pwsh -File`, omitting the runner's `exit $LASTEXITCODE` wrapper — so no module can pin a step exit code, and two already dobugtestingagent-ready | unassigned | yesterday |
| #1172 | L214 claims pwsh shifts the argument binding on an unset $env: reference; its own evidence says the binder refuses. Three docs and 17 call sites inherit the wrong modelbugdocumentationagent-ready | unassigned | 2 days ago |
| #1171 | Status feed: pending_gates counts runs AT a gate, not runs stopped BY one (703 holds 2, page says 1)bugagent-ready | unassigned | 2 days ago |
| #724 | Migrate remaining value out of FFC-IN-AI-Management into the hub (inventory + first tranche)documentationagent-ready | unassigned | 2 days ago |
| #1165 | CI guard: fail on a tracked filename with a non-printable or private-use characteragent-ready | unassigned | 2 days ago |
| #912 | 228 has never succeeded: a secrets.* reference against an environment with no secrets, and nothing catches the classsecurity | unassigned | 2 days ago |
| #1152 | 740 records only the run conclusion, so a failure that changes cause reads as the same outage continuing — 735 moved to the credential step and both rolling issues hid itbugagent-ready | unassigned | 2 days ago |
| #1084 | 728 never runs in the merge group, so a red hooks suite can land on main — measured on #1018 + #1039bugblocked | unassigned | 2 days ago |
| #894 | Fleet security headers: which FFC sites actually serve themsecuritypriority: high | unassigned | 2 days ago |
| #841 | Fleet security-audit coverage gap: Node repos with no vulnerability detectionsecuritypriority: high | unassigned | 2 days ago |
| #1033 | 🚨 Scheduled workflow failing: 735. Repo - Dependabot Affected Repos [Org]bug | unassigned | 2 days ago |
| #893 | Fleet smoke engine drift detectedpriority: highsmoke-failure | unassigned | 2 days ago |
| #848 | Key Vault: read-all-cbm-github-pat is dead (502's deliver job down 3 days), and the read/write split is unenforced (per-secret RBAC + liveness monitoring)securitypriority: high | unassigned | 2 days ago |
| #1048 | 703's weekly refresh has silently missed 2 runs: both monitors filter status=completed, so a schedule that stops reads greenbugagent-ready | unassigned | 2 days ago |
| #983 | dry_run is enforced per-step, not per-job: every dry run of 701/702/720 parks on a write gate and mints a write credentialsecurityblocked | unassigned | 3 days ago |
| #1028 | App-identity pushes fire no pull_request event: the PR keeps the previous commit's green checks and reads as CI-verifiedbugagent-readyblocked | unassigned | 3 days ago |
| #1135 | Nothing observes the pickup query an agent actually types: #1028 sat in the agent-ready results for 3 days while explicitly blockedbugagent-ready | unassigned | 3 days ago |
| #1133 | 703 is the last scheduled workflow on a reviewer gate, and 64 runs have declined it because there is nothing to review (#834 tail)priority: highagent-ready | unassigned | 3 days ago |
| #1070 | 745 board audit is blind to agent work that was never labelled: expected comes from a hand-applied labelbugagent-ready | unassigned | 3 days ago |
| #1111 | Board audit cannot see work that merges before it is carded: #1108 merged uncarded and every category reported 0bugagent-ready | unassigned | 3 days ago |
| #1042 | The Conductor runs unguarded: .claude/hooks/ is never loaded for the workspace session rootsecurityagent-ready | unassigned | 4 days ago |
| #1072 | Bash-tool commands collapse every doubled backslash before the shell: six runs blamed the heredoc, and the surface is the command stringbugagent-ready | unassigned | 4 days ago |
| #1027 | Hook rule tests have no per-rule polarity check: a rule tested only in the direction it passes is green and wrong (4th instance)bugclaimedagent-ready | unassigned | 4 days ago |
| #788 | Fleet: upload-pages-artifact@v5 silently drops dot-entries — .well-known/security.txt 404s on affected sitesagent-ready | unassigned | 4 days ago |
| #1100 | Nothing detects a Dependabot advisory deadlock: 2 PRs each fixing one high advisory kept the main website red for 4 days, and neither could ever go greensecuritypriority: highagent-ready | unassigned | 4 days ago |
| #880 | Every FFC Pages deploy silently discards next.config.ts (configure-pages writes its own .js)bugpriority: highagent-ready | unassigned | 4 days ago |
| #1119 | test(workflow-logic): 104 of 1170 tests fail on Windows while ubuntu-only CI stays green | unassigned | 4 days ago |
| #1040 | Hook rule coverage is blind to the warn() channel: #1018's two rules land green with zero casesbugagent-ready | unassigned | 4 days ago |
| #755 | Canary drill: prove the smoke engine catches each failure class it claims (child of #752) | unassigned | 4 days ago |
| #934 | Smoke never probes www: the engine only mentions it in an error hint, while 13 of 55 live sites are broken on wwwbugpriority: high | unassigned | 4 days ago |
| #1099 | Conductor PR supervision is hub-only by accident: 79 open PRs across the four satellite repos, oldest 6.5 months, never sweptblocked | unassigned | 5 days ago |
| #1109 | test_powershell_command_resolution: one test does not guard on host absence — one false red, and two greens that measure nothingbug | unassigned | 5 days ago |
| #936 | Fleet sync of the www smoke probe will red-line 13 sites at once: SMOKE_REQUIRE_WWW is unset everywherebugpriority: highagent-ready | unassigned | 5 days ago |
| #1103 | Public status feed drops a whole repo's PRs when its last agentic-os issue closes: freeforcharity.org's 22 open PRs vanished 6 minutes after a security fix landedbugagent-ready | unassigned | 5 days ago |
| #823 | WHMCS: auto-populate client custom fields from product-order answers (order → client sync)enhancement | unassigned | 6 days ago |
| #1096 | A close/reopen round trip launders a `claimed` label past both release paths: #986 sat claimed for 4 days with no open PRbugagent-ready | unassigned | 6 days ago |
| #986 | Six orphan branches have no open PR and an unknown disposition — decide by content, not by ancestrydocumentationblocked | unassigned | 6 days ago |
| #1091 | 727's behind-count counts merge commits and doc-only churn: two ledger PRs red-line every open PRbugagent-ready | unassigned | 6 days ago |
| #1087 | The ledger's guard-path check proves existence, not enforcement — a row can name a guard its own lesson says is incompleteagent-ready | unassigned | 7 days ago |
| #1024 | Nothing checks that a doc's "Tracked on #N" pointer is still open — 3 of 4 cite closed issues, and one was telling agents to discard real evidenceagent-ready | unassigned | 7 days ago |
| #1085 | Readiness audit models only 1 of 727's 2 fail conditions, so it reports ok on PRs the guard will reject (measured on #1082)bugagent-ready | unassigned | 7 days ago |
| #863 | Establish the canonical URL standard on FFC_Single_Page_Template (trailingSlash + sitemap + canonicals + lighthouserc)enhancementagent-ready | unassigned | 7 days ago |
| #835 | The public Agentic OS board has no auto-add: every Conductor run adds items to Project #9 by handenhancementagent-ready | unassigned | 7 days ago |
| #1078 | Open-PR readiness audit is hub-only: 8 of 13 open agentic-os PRs are unwatched, 2 queue-blocked for 9 daysagent-ready | unassigned | 7 days ago |
| #1074 | 229 + 230: split the dry-run preview onto whmcs-prod-read so a rehearsal needs no gate | unassigned | 7 days ago |
| #1026 | guard_bash.py rule numbers are a shared namespace with no uniqueness check: 6 and 7 are each claimed twice across 4 open PRsbugagent-ready | unassigned | 8 days ago |
| #1068 | 101 Domain Status reports nothing when there is something to report: a tolerated DKIM warning still fails the step, skipping summary and post_backbugagent-ready | unassigned | 8 days ago |
| #1067 | 737 claim-sync reads only the PR body, so a Refs in the title claims nothing — #823 sat pickable while #825 implemented itbugagent-ready | unassigned | 8 days ago |
| #1065 | A waiting approval gate can be running an unmerged PR branch, and nothing surfaces the ref: the safety doc classifies workflows as they are on mainagent-ready | unassigned | 8 days ago |
| #1060 | pwsh invocation guard skips a call whose script path is a variable: 701's 11-parameter content call is unjudgedbugagent-ready | unassigned | 8 days ago |
| #1057 | Dependabot has capped out in 35 of 62 fleet repos: 793 open PRs, nothing merges them, so Actions updates have stopped being filedpriority: highagent-ready | unassigned | 8 days ago |
| #1006 | candid-prod-read does not exist as a GitHub environment: 801/802 have never been runnable, and it blocks #1005security | unassigned | 8 days ago |
| #1044 | Nothing validates that a sites-list Repo URL resolves: a 404 sat in the public dataset until another repo's link checker found itbugagent-ready | unassigned | 9 days ago |
| #1023 | test_729 runs the step in the checkout, loses all 5 of its tests, and leaves a stray file — while CLAUDE.md points at closed #945bugagent-ready | unassigned | 9 days ago |
| #1022 | `$?` guard misses the `&&` and `||` separators: `cmd | tail && echo $?` still prints a false EXIT=0bug | unassigned | 9 days ago |
| #1037 | 111's apply job always fails after succeeding: curl.exe does not exist on ubuntu-latest, and the only assertions that could fail are exit-0 warningsbugagent-ready | unassigned | 9 days ago |
| #1035 | 734 reaps a gate on its first visible tick when the run was queued behind a concurrency group: age is measured from created_at, not from when the gate openedbugagent-ready | unassigned | 9 days ago |
| #771 | Standardize on pnpm across FFC templates and repos (migrated from FFC-IN-AI-Management#9) | unassigned | 9 days ago |
| #1031 | ffcadmin's two src/data + public/data pairs are not the same case: guard workflow-catalog.json, delete the dead agentic-os-status.json copybugagent-ready | unassigned | 9 days ago |
| #1030 | 734 publishes an exact reap deadline that has been wrong on 26 of 26 runs: the cron instant is not the delivery instantbugagent-ready | unassigned | 9 days ago |
| #1019 | Text-scanning guards flag prose about the thing they catch — four instances in 48h, and the newest rule shipped with the bugbug | unassigned | 10 days ago |
| #1016 | sites-list ordering: group stable GitHub Pages sites on top, in the generator (recovered from #425 before its branch is deleted)enhancement | unassigned | 10 days ago |
| #908 | ffcadmin data-refresh PRs hang forever once they fall behind main — three public dashboards went stale for 6h with all checks greenbug | unassigned | 10 days ago |
| #771 | Guard that the dual-written data feeds cannot drift (public/data vs src/data)agent-ready | unassigned | 11 days ago |
| #834 | Reads-level GitHub workflows are gated on github-prod: 8 of 21 scheduled runs failed or were cancelled waiting for approvalagent-ready | unassigned | 12 days ago |
| #860 | Standardize on Node 24, and make the templates the rehearsal surface for every fleet-wide changeenhancementpriority: high | unassigned | 12 days ago |
| #729 | Repo hygiene for the cutover test target FFC-EX-technologymonastery.org (child of #726) | unassigned | 12 days ago |
| #859 | Baseline compliance is unmeasured across the fleet — policy pages, consent and security must hold regardless of template divergenceenhancementpriority: highagent-ready | unassigned | 12 days ago |
| #748 | Conditional basePath missing: default-Pages-URL sites serve pages with 404 assets (technomonasteries, Footer_Only_Template; audit fleet)agent-ready | unassigned | 12 days ago |
| #918 | www is unverified across the fleet: aprilhansen.com serves 526 on www, and both the cutover and the smoke reported OKbugpriority: high | unassigned | 13 days ago |
| #756 | 2 high Dependabot alerts survive both open dependency PRs: fast-uri@3.1.2 needs a pnpm override, not a bump | unassigned | 13 days ago |
| #865 | No per-charity traffic data — the traffic-ordered rollout has nothing to order byenhancement | unassigned | 13 days ago |
| #883 | First working telemetry sweep: findings, and why the GA4 mismatch check is unreliable | unassigned | 17 days ago |
| #844 | Credential consistency: 4 environments still hold raw secrets in GitHub instead of Key Vaultsecuritypriority: high | unassigned | 17 days ago |
| #864 | Search Console + Analytics baseline on the two FFC sites — the proof points for the URL standardenhancement | unassigned | 18 days ago |
| #861 | Epic: prove the web standard on FFC first, then propagate to the fleet in order of trafficenhancementpriority: high | unassigned | 18 days ago |
| #838 | Security: 34 of 44 Node FFC-EX repos have no vulnerability detection at all (why #822's scope read 13 instead of 38) | unassigned | 18 days ago |
| #822 | Security: bump Next.js to patched 16.2.11 across the fleet (scheduled Security Audit red on both templates)enhancement | unassigned | 19 days ago |
| #813 | Epic: WHMCS fraud review via FraudLabs Pro + daily agent check (charities mis-flagged as fraud)enhancement | unassigned | 19 days ago |
| #752 | Epic: process assurance — verify the Agentic OS's own processes keep working (monitor the monitors) | unassigned | 19 days ago |
| #816 | Enhance the Charity Website product request (pid 40) to collect full-template content + self-hiding sectionsenhancement | unassigned | 20 days ago |
| #770 | Build a custom WHMCS MCP server (migrated from FFC-IN-AI-Management#3) | unassigned | 20 days ago |
| #778 | Consolidate Cloudflare record-writing: one engine for the apex/www Pages standard + shared API libenhancement | unassigned | 21 days ago |
| #744 | Publish FFC Zeffy campaign list for smoke auto-classification (charity-specific vs ffc-interim donations) | unassigned | 22 days ago |
| #667 | Bot data PRs stuck at CI 'action_required' again despite approval policy fix (#637 regression, different cause) | unassigned | 22 days ago |
| #670 | Weekly Linkinator (external) fails on synthetic issues/new links: skip pattern + 2 stale repoUrl rows | unassigned | 23 days ago |
| #573 | Add session/run budget guidance to governance (Max subscription capacity) | unassigned | 24 days ago |
| #769 | Build a custom Zeffy MCP server (migrated from FFC-IN-AI-Management#2) | unassigned | 24 days ago |
| #574 | [ops-concierge] Autonomy pilot — tracking issueops-concierge | unassigned | 24 days ago |
| #747 | Fleet-status donation badges + campaigns dashboard page on ffcadmin (child of #745) | unassigned | 24 days ago |
| #745 | Epic: charity campaign registry — capture and track every Zeffy campaign across FFC-supported sites, surfaced in the dashboards | unassigned | 24 days ago |
| #738 | Epic: unify fleet smoke patterns into one Agentic-OS-integrated verification system with public visual fleet status | unassigned | 24 days ago |
| #638 | Lighthouse CI failing on main: performance 0.67 vs 0.9 budget (bootup-time, CLS, DOM size, contrast) | unassigned | 25 days ago |
| #572 | Quarterly inventory refresh + reflection session | unassigned | 25 days ago |
Latest Conductor log
## Run 152 — START (2026-08-12, ~04:0xZ) · core 4983 / graphql 4905 Conductor run beginning. Workspace bootstrapped: all five core clones fetched, `symbolic-ref` checked per clone (L-run-149 addendum), **all five on `main`, 0 behind, 0 dirty** — no stranding this run, unlike runs 145 and 149. **Run number derived from this issue's tail** (`--paginate`, streaming `--jq`), newest START = 151 → 152. Carried into this run from #151's END: 1. **#1190** (mechanical half of L244 — `URLError` bounde… (truncated)
## Run 151 — END (2026-08-12, 00:4x–01:5xZ) · core 4979 → 4982 / graphql 4920 → 4950 **2 PRs merged (#1191 mine, plus #1189 recovered from run 150) · 1 delivery merged and verified live (ffcadmin #910, delivery 85) · 1 issue filed (#1190) · 1 ledger row (L244) · 1 confirmed review finding posted to #1127 · 0 gates approved (87th hold on 703) · 4 board card ops, audit rc 0 on the first pass · no Clarke contact.** ### Thread 1 closed — and the cause was not the one the thread expected #1189 did… (truncated)
## Run 151 — START (2026-08-12, ~00:4xZ) · core 4979 / graphql 4920 Conductor run 151. Newest START in this log was run 150 (`2026-08-11T22:04:20Z`), so this is 151. **Step 0 bootstrap complete.** All five core clones fetched, on `main` via `symbolic-ref`, `0` tracked-file changes each — none stranded on a merged branch this run (run 149's addendum found two, run 145 found the same two). Heads: hub `58138b3`, freeforcharity `88593b3`, ffcadmin `b3cd374`, Single_Page_Template `edfd3ff`, Footer_… (truncated)
## Cloud worker — landing run, 2026-08-12 **Step 1 gate fired:** 3 open `agentic-os` PRs (#1189, #1127, #1039) → landing run, no new work opened. ### Assessment of the queue All three were already CI-green, `mergeable_state` clean, and **free of unresolved review threads**. Nothing to fix in the usual sense: | PR | State | Threads | Landable by an agent? | | --- | --- | --- | --- | | #1189 | non-draft, green, at `main`'s tip | 0 | Conductor's docs PR — yours to enqueue | | #1127 | draft, gre… (truncated)
## Run 150 — END (2026-08-11, 22:0x–22:4xZ) · core 4991 → 4996 / graphql 4942 → 4958 **1 agent PR reviewed and landed (#1187, `58138b3d`) · 1 delivery merged and verified live (ffcadmin #909, delivery 84) · 1 ledger PR opened and sitting in the merge group (#1189, L242/L243) · 1 issue filed (#1188) · 0 gates approved (86th hold on 703) · 4 board card ops, audit rc 0 on the first pass · no Clarke contact needed.** ### Step 2 — gates: one, held `31370775983` (**703. Generate Sites List**) `wait… (truncated)
## Run 150 — START (2026-08-11, ~22:0xZ) · core 4991 / graphql 4942 Run number derived from this issue's tail (page 6, newest = `Run 149 — END` 20:07:25Z), not from local notes. **Step 0 bootstrap:** all five core clones fetched, `symbolic-ref` checked per clone (run 145/149 rule), all on `main`, all clean. - hub `5baee0d` · freeforcharity.org `88593b3` · ffcadmin.org `b7fc496` · Single_Page_Template `edfd3ff` · Footer_Only_Template `d163883` **Carried threads under test this run (from run 14… (truncated)
## Run 149 — END (2026-08-11, 19:0x–20:0xZ) · core 4995 → 4990 / graphql 4944 → 4856 **3 PRs merged (#1183, #1184, #1186) + 1 delivery merged and verified live (ffcadmin #907, delivery 83) / 3 review findings fixed on my own PR / 1 issue filed (#1185) / 2 ledger rows / 0 gates approved (85th hold on 703) / 5 board card ops, audit rc 0 on the first pass / no Clarke contact needed.** ### The run's finding: run 148 reported supervision complete while three findings sat unread on its own draft Ru… (truncated)
## Run 149 — START (2026-08-11, ~19:0xZ) · core 4995 / graphql 4944 Run number taken from #719's tail (newest START = 148, ENDed 16:41:44Z), not from `state/CONDUCTOR.md` — which this run found has **no run-148 outcomes section at all**, only addenda. The file's own rule held. **Bootstrap: two of five core clones were stranded on already-merged branches** — hub on `conductor/l238-l239-async-fields-and-prior-art`, ffcadmin on `data/agentic-os-status-delivery-82`. Same finding as run 145, same t… (truncated)
## Cloud worker — landing run (no new work opened) **Step 1 tripped:** 4 open PRs labelled `agentic-os` (#1039, #1127, #1183, #1184) ≥ 3, so this run went to landing rather than new work. No new branch, no new PR, nothing pushed. ### State of the four | PR | CI | Threads | Actually blocked on | | --- | --- | --- | --- | | #1183 | all green | **3 unresolved** | author fix — see below | | #1184 | all green | **1 unresolved** | author fix — one-char | | #1127 | all green | 1, resolved | **Clarke… (truncated)
## Run 148 — END (2026-08-11, 16:0x–16:5xZ) · core 4986 → 4997 / graphql 4922 → 4914 **1 PR merged (#1181) + 1 delivery merged (ffcadmin #906, delivery 82) + 2 shipped as drafts (#1183, #1184) / 0 issues filed / 2 ledger rows / 0 gates approved (84th hold on 703) / 4 board corrections, audit rc 0 → rc 0 / no Clarke contact.** ### The run's finding: the metric built to notice unlanded work was reporting zero 739's `readyWaiting` — the #900 metric, added *because* every other signal reads healt… (truncated)