dogfood
Skill2 starsChanged 46 days agoMITView on GitHub
Dev-only, anti-hall repo (not part of the shipped plugin). Use in every anti-hall development session. The session that builds anti-hall also runs it - observe every anti-hall message, telemetry, engine health and bug signal, collect reports from other projects' running sessions, log every misfire with evidence, and drive fixes.
81 other projects use a version of this.
Adopted by
Most recently taken up by
- rodos4312/Meu-Agenttheir own version, first seen Nov 15, 2026
- getlien/lientheir own version, first seen Nov 15, 2026
- IgorGanapolsky/mac-yolo-safeguardstheir own version, first seen Nov 15, 2026
- axiom-desu/ZCodiumtheir own version, first seen Nov 15, 2026
- lsjwzh/MultiCCtheir own version, first seen Nov 14, 2026
- NousResearch/hermes-agenttheir own version, first seen Nov 13, 2026
Used with
Projects with it also use dep-refs, feature-boundary-planner and 10 more.
Other files people use alongside this one.
Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
axiom-desuSkilldep-refs
Use when needs to inspect TypeScript export references in the z-code workspace, list exports from a file, verify whether an export is unused before deletion, investigate who imports a symbol during refactors, or combine pnpm knip unused-export results with pnpm dep:refs symbol-level reference tracing.
Also in 6 projects that use this
axiom-desuSkillfeature-boundary-planner
Map a ZCode behavior change to current UI surfaces, state owners, protocol commands, persistence, and validation. Use for impact analysis, product-boundary planning, or an implementation handoff grounded in the checked-out source.
Also in 6 projects that use this
lir-navcooSkillyuanbao
Yuanbao (元宝) groups: @mention users, query info/members.
Also in 18 projects that use this
Cartograf666Skillarchitecture-governance
Bootstrap, document, enforce, and review a project's architecture through layer responsibilities, dependency direction, architecture manifests, legacy baselines, Change Envelopes, cross-layer approval policy, and deterministic architecture checks. Use when starting a project, auditing an existing codebase, planning cross-layer work, preventing architecture erosion, reviewing structural changes, or deciding whether implementation may expand into another layer.
Also in 6 projects that use this
math-incSkillminecraft-modpack-server
Set up a modded Minecraft server from a CurseForge/Modrinth server pack zip. Covers NeoForge/Forge install, Java version, JVM tuning, firewall, LAN config, backups, and launch scripts.
Also in 5 projects that use this
math-incSkillpokemon-player
Play Pokemon games autonomously via headless emulation. Starts a game server, reads structured game state from RAM, makes strategic decisions, and sends button inputs — all from the terminal.
Also in 5 projects that use this
- +1 more from math-inc
Other versions81 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 81
- NousResearch/hermes-agent★ 264ksame name
- vercel-labs/agent-browser★ 51ksame name
- coder/coder★ 17ksame name
- zai-org/ZCode★ 8.7ksame name
- callstackincubator/agent-device★ 5.4ksame name
- callstack/agent-device★ 5.4ksame name
- salesforce/cloudsplaining★ 2.3ksame name
- coder/cmux★ 2.1ksame name
- coder/mux★ 2.1ksame name
- phodal/routa★ 1.8ksame name
71 more not listed
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository7 other files it gives its agents.
- AGENTS.mdAGENTS.md
- llms.txtllms.txt
- engine-debugSkill
- engine-laneSkill
- gh-workSkill
- releaseSkill
- repo-hygieneSkill
The file
We split this skill into the rules it gives an agent
2 rules. 0 are common, 2 are distinctive, and 0 are very specific to this project.
0
Common
Most of the 81 edited versions we read keep these.
2
Distinctive
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 53
Every progress report to the owner includes: open entries by severity, fixed since last report, mistake rates of acting features.
- Distinctive: line 98
Never close or delete an issue automatically.
“Common” is measured: kept by at least 6 in 10 of the 81 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Dogfood anti-hall while developing it
Dev-only: this skill lives in the repo's .claude/skills/, outside plugins/anti-hall/, so it is
never installed for users (like a devDependency). Every anti-hall signal in a dev session is test
output. Judge it, record it, fix it. Never silently obey, dismiss or work around one.
1. Every anti-hall message, as it arrives
Blocks, advisories, nudges, tracker lines, stop-hook feedback, statusline segments, DevSwarm
notices. Verdict: true positive / false positive (wrong about facts) / noise (right
but useless or repeated) / wrong wording. Anything but a true positive gets one line in
.anti-hall/history/dogfood/ISSUES.md (local, gitignored):
- YYYY-MM-DD HH:MM | <check / event / source> | <verdict> | <message quote + contradicting fact> | open
A repeat of an open entry gets its own dated line (frequency is data). Keep working.
2. Telemetry and health (hourly, at every milestone, and before ending the session)
sh ~/.anti-hall/ah-engine-live/status.sh: errors, panics, restarts, breaker, crash-loop, daemon RSS.sh ~/.anti-hall/ah-node-shadow/node-shadow.sh --compare: any check where the engine is weaker than Node is a P0 entry; defer counts per check are tracked over time.ah-engine devswarm ledger --since 1h(andah-engine telemetry): per-feature actions, success rate, refusals, mistake rate. A rising mistake rate or a new refusal reason is an entry.- The engine event log: whole-event fallbacks, stuck workers, script "interrupted"/CPU-limit errors, git timeouts. Record count and check id.
- The repo itself: a stale
.git/index.lock, temp-dir growth (ls $TMPDIR | wc -l), leftoverah-engine serveprocesses, disk growth. Each is a bug signal until proven otherwise.
3. Other projects' running sessions (every few hours and at milestones)
anti-hall runs in every project on this machine. List live sessions (ListAgents, peer
sessions with status busy/idle, not offline) and send each a short message asking for
anti-hall misfires, wrong blocks, slowdowns or confusing messages they have hit since the last
ask, with the exact message text and what they were doing. Ask; never instruct them to change
anything. Record each answer (or "no reply") as entries with source peer:<session name>.
Owner-ratified peer channel; keep it to one message per session per round. DevSwarm
workspace sessions are never messaged directly (devswarm-comms-guard blocks it, correctly):
ask that project's Primary to collect from its workspaces and relay.
4. Fixing
- Group open entries per check into one lane: confirm the cause from code/data, sibling sweep, regression test proving the true-positive case still fires, then the fix (plugin config/JS or engine; Node gets bug fixes only and stays in parity).
- Mark entries
fixed <sha>/wontfix <reason>; never delete entries. - Every progress report to the owner includes: open entries by severity, fixed since last report, mistake rates of acting features.
5. All work lives on GitHub (issues + the "anti-hall roadmap" board)
Tasks, bugs, features, plans and the roadmap live in the repo's issues and on the user project
board "anti-hall roadmap" (project 3, owner talas9; views Now, Next, Backlog,
By milestone, Board). The session task list is built FROM the board, never the other way
round. Exact commands, field ids and option ids: the gh-work skill.
Setup this relies on
- Issue forms (
.github/ISSUE_TEMPLATE/): Bug report, False positive, Feature request (blank issues off). Their Priority / Estimate / Area / "Estimate in hours" fields feed the automation. - Labels:
type:feature|bug|docs|chore,priority:P0..P3,size:S|M|L|XL,area:*(engine, devswarm, hooks, statusline, codex, ci, docs, release, plus the guard areas),status:triage|accepted|in-progress|blocked. triage.yml: maps form fields topriority:*/size:*/area:*, addsstatus:triageon open, labels PRs by path (labeler.yml), and adds new issues to the board when its token secret is set (else the board's own auto-add).community.yml: rules triage + one model brief comment. The moderation workflows (when present) addmoderation:review,needs-issue,risk:*and triage suggestions: a session acts on them (seegh-work).- Milestones: v0.300.0 (engine decides hooks, Node is the fallback), v0.301 engine-only (Node removed), v1.0 (no Node; everything goes through the engine; contract frozen).
Operating loop (autonomous)
- Session start / resume: sync the board (
gh project item-list 3 --owner talas9), read the Now view, then Next by priority. Each session task subject starts with#n. - Anything new (an owner request, idea, bug or plan, or something found while working): open the issue FIRST, from the form fields, and triage it at once (labels, board fields, milestone). Plans go in the issue body or a linked comment; a local plan file links its issue.
- Pick the next work yourself: the highest-priority Now item with no in-flight agent, else
the top of Next. Move it to In progress (
status:in-progress+ board Status), comment the approach, work it, comment the evidence, and close it through a PR withCloses #n. Never wait for permission on accepted items. Stop only at owner-only gates (billing, secrets, destructive actions): comment on the issue, label itstatus:blocked, take the next item. - Every few hours (with the health round of section 2): re-sync the board, triage anything
new (bot suggestions included), post a one-line progress comment on each in-progress issue,
and handle
status:in-progressissues silent for 24 h andstatus:triageissues older than a day.
Rules
- Triage =
status:accepted+ priority/size/milestone, or a comment saying why it is deferred. Never close or delete an issue automatically. - One comment per milestone (approach, lane/commit shas, test or replay evidence, blockers), not per step.
- Branch
<type>/<n>-slug, commits reference#n, PR bodyCloses #n, so a merge moves the item to Done. - Done: the closing comment states what was verified and how (command and result). Work shipped without proof stays open with the missing check named.
- Dogfood entries that need a code fix get an issue (
type:bug, area of the check), linked from the ISSUES.md line (issue #n). - Issue text stays agnostic: no private paths, names or session ids (public repo).
Rules that still apply
Evidence before verdicts. Never disable a guard to get past it; log it and work within it. Safety switches change only with the owner's OK.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

