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/watt-mind/factory/factory-auditgit clone --depth 1 https://github.com/watt-mind/factoryWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/commands/watt-mind/factory/factory-audit)<a href="https://agentmods.dev/commands/watt-mind/factory/factory-audit"><img src="https://agentmods.dev/badge/commands/watt-mind/factory/factory-audit.svg" alt="Measured on agentmods" height="20"></a>What 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 | $0.00015 | $0.00916 |
| Opus 5 | $0.00008 | $0.00458 |
| Sonnet 5 | $0.00003 | $0.00183 |
| Haiku 4.5 | $0.00002 | $0.00092 |
Grade A, and why
factory-audit 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 4d ago.
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 — 40 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Audit a repository against the conventions baseline in ~/Develop/hdkiller/docs/guides/project-conventions.md and turn the gaps into tracked work.
Default target is the repo I'm currently in. Interpret $ARGUMENTS as a different repo path, specific check IDs to run in isolation, or all (every repo in config/repos.yaml that exists locally — audit each and report a matrix).
1. Grade every check
Work through the §7 conformance table — PC-01 … PC-20 — in order. That table is the source of truth for what's required; read it fresh each run rather than from memory, since it changes as the baseline grows.
For each check, record PASS / FAIL / N/A plus the evidence you actually observed: the command you ran and its relevant output, or the file and line. An unverified PASS is worse than a FAIL — it certifies something nobody checked.
Rules that keep the grade honest:
- Verify against the live repo and the live GitHub settings, not against what
AGENTS.mdor this guide claims.PC-20exists precisely because docs drift; when a doc and reality disagree, the doc is the finding. "Live" meansorigin/<base>, sogit fetch --quietfirst and follow the floor's Checkout freshness rule — a FAIL graded against a checkout that predates the fix is a fabricated finding, and it gets filed as a ticket for work already done. - N/A requires a stated reason grounded in the repo ("no auto-deploy branch", "free private plan blocks branch protection"). Never N/A a check because it's inconvenient or because fixing it is out of scope — that's a FAIL with a filed ticket.
PC-03(runner policy) andPC-18(missing secrets) are cheap and high-consequence — on private repos cloud runners bill real money; on public repos, fork PR sandboxing requires standard hosted runners or ephemeral isolation, while self-hosted remains for trusted internal lanes. Always check them, even in a partial audit.- For
PC-17, a workflow that exists but is disabled by a staleif:gate is a FAIL, not a PASS — a gate whose stated precondition has since been met is the most common way a repo loses its strongest test signal without anyone noticing. - Read-only. Do not fix anything during grading; grade first, so the report reflects the repo as found.
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.
- 4d ago First seen · 40 lines · 0 tokens per session scan A 68f87be9a5d1
factory-audit is a command published in the GitHub repository watt-mind/factory (10 stars, last pushed 4d ago), licensed Apache-2.0. It adds 15 tokens to every session and 916 once invoked, about $0.0001 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 2026-08-31.
Other commands, from other repositories
release
Prepare, cut, and verify a warren release — tracker audits, version bump, CHANGELOG curation, ROADMAP update, push, then watch the pipeline through to published artifacts.
app-build
Run the sprint — launch developer agents in parallel, review each PR, gate to QA, surface bugs back into the loop.
app-run
Mostly-autonomous end-to-end run — idea → scope-lock → sprint loop → ship-readiness, surfacing only blockers and the two human gates (scope-lock, ship).
app-status
Show the current state of the project — vision, sprint goal, board, blockers, daily report.
app-audit
Grade an existing app against the Mobify Studio House KB and Axiom standards — fan out the specialist auditors, produce a severity-ranked gap report, build a remediation backlog, then fix (safe fixes automatically, risky changes only with your approval).
app-plan
Plan the next sprint — tech-manager turns backlog + impl specs into a parallelizable board.