Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add commands/jayminwest/warren/releasegit clone --depth 1 https://github.com/jayminwest/warrenWhat it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00039 | $0.01293 |
| Opus 5 | $0.00019 | $0.00647 |
| Sonnet 5 | $0.00008 | $0.00259 |
| Haiku 4.5 | $0.00004 | $0.00129 |
Grade A, and why
release scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured yesterday.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 120 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Release
Prepare a new warren release, push it, and verify the pipeline shipped it. The release does not exist when the commit lands — it exists when the Release workflow's five-job chain finishes: gates → tag → draft release → ghcr image + GKE deploy → npm publish → promote draft to final → Discord announce. This command runs everything up to the push, then watches that chain to the end.
1. Pre-flight
- Working tree is clean, on
main, up to date with origin (git status,git pull). - No open PRs about to auto-merge into the release commit:
gh pr list --state open. If one is close to merging, either wait for it or release without it deliberately — never race it. - Find the last release tag:
git describe --tags --abbrev=0.
2. Tracker audits (two parallel subagents)
Spawn both subagents in a single message so they run concurrently. Neither pushes — all commits ride out with the single push in step 9.
- Seeds audit — a subagent that invokes the
seeds-issue-auditskill and follows it exactly: auto-close only HIGH-confidence issues with citable evidence, runsd syncafter closing, report borderline cases, never push. Have it return its two tables (auto-closed / borderline). - GitHub issue sweep — a subagent that lists open GitHub issues
(
gh issue list --state open) and verifies each against HEAD. Same discipline as the seeds audit: close an issue only when the fix is verifiably onmain(name the commit or merged PR in the closing comment,gh issue close <n> --comment "..."); anything uncertain goes in a borderline table, untouched. Staleness alone never closes anything.
Hold both reports — the borderline tables go in the final summary (step 10), and confirmed-done plans feed the ROADMAP update (step 6).
3. Review changes since the last tag
git log <tag>..HEAD --onelineandgit diff <tag>..HEAD --stat.- Determine the version bump level. Always use patch unless the user explicitly requests minor or major.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- yesterday First seen · 120 lines · 39 tokens per session scan A ed35ca3425a4
release is a command published in the GitHub repository jayminwest/warren (346 stars, last pushed 3d ago), licensed MIT. It adds 39 tokens to every session and 1,293 once invoked, about $0.0002 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other commands, from other repositories
factory-merge
Review open PRs thoroughly, fix what's fixable, merge what's good.
factory-retro
Find what is repeatedly wasting the factory's time and fix the harness, not the symptom.
factory-ticket
Implement exactly one already-claimed Linear ticket in the current worktree.
factory-friction
File harness friction observed in an interactive session — capture only, never implement.
factory-next
Recommend and optionally run the next factory stage for this repo.
factory-report
Linear status report across configured repos — queue depth, triage backlog, ready-to-dispatch, blocked/held.