Getting it into your agent
It runs from inside its repository, so the clone comes first — what it calls does not travel with the file alone.
git clone --depth 1 https://github.com/OneDro1d/dark-factorynpx agentmods add commands/onedro1d/dark-factory/validateWrote 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/onedro1d/dark-factory/validate)<a href="https://agentmods.dev/commands/onedro1d/dark-factory/validate"><img src="https://agentmods.dev/badge/commands/onedro1d/dark-factory/validate/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/commands/onedro1d/dark-factory/validate"><img src="https://agentmods.dev/badge/commands/onedro1d/dark-factory/validate.svg" alt="Reviewed on agentmods" width="80" 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 5.1 | $0.00019 | $0.05358 |
| Opus 5.5 | $0.00008 | $0.02143 |
| Sonnet 5.5 | $0.00004 | $0.01072 |
| Haiku 4.5 | $0.00002 | $0.00536 |
Grade C, and why
validate scanned grade C with 1 finding 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 21d 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.
Recursive force deletehighDestructive command
rm -rf with a variable or a broad path is one typo away from removing the wrong tree.
rm -rf ../workers/dev/*probe* 2>/dev/null || true How it starts
The opening of the file, as written. The whole thing — 316 lines — stays where its author put it; the contents beside it link to each section on GitHub.
You are validating a freshly installed Dark Factory kit. This session's cwd is a throwaway
notepad that validate-arm.sh armed for exactly this run — do not go looking for "the
working notepad" the way VALIDATE-INSTALL.md Part 1's Task 0 does; you are already in it,
and M-VALIDATE is already RUNNING in .df/missions/M-VALIDATE/state. The kit root is
the parent of your cwd (..). The record is the one that RESOLVES for THIS machine, not
"the file at the kit root": if LOOM_LOCK is set in your environment, that is it; otherwise
run bash <kit-root>/vendor/dark-factory/boot-kit/scripts/identify.sh --match <kit-root>/instances
(fall back to <kit-root>/boot-kit/scripts/identify.sh, if this kit ships one) and use the
record it names, else the single <kit-root>/*.lock.json. A non-zero exit from --match now
means no record matches this machine, not an error. On every Coder in a shared instance repo
the kit-root file describes a DIFFERENT machine (the laptop) and identify.sh exits 3 against
it — a run that validated that file would be the exact failure Task 1 exists to catch. Quote
which record you used and how it was chosen.
Work the tasks below in order. Record each as PASS, FAIL, or UNKNOWN — UNKNOWN means you could not probe it, not that it passed. Quote commands and output verbatim; never paraphrase a denial or an error. Fix nothing until the end.
⛔ NEVER PRINT A SECRET — and a whole-environment dump IS printing one. This session's
transcript is written to disk and kept. On a Coder workspace the pod env carries live hub PATs
and the agent token, and ~/.mcp.json can hold literal bearers. MEASURED on the homelab box
(2026-09-11 and twice on 2026-09-18): three validate runs each opened with a bare env, and
one ran claude mcp get onedroid. Every one of those runs copied a live token into its
transcript. So:
- Do not run
env,printenvwith no argument,set,export -p,declare -x,claude mcp get …, orcat/jqon~/.mcp.json,~/.claude.jsonor any rendered MCP config. - Read one named, non-secret variable at a time, e.g.
printenv LOOM_LOCKorprintenv CLAUDE_CODE_ENTRYPOINT. To learn whether a secret is SET, test it without echoing it:[ -n "${SYNAPSE_ONEDROID_PAT:-}" ] && echo set. - For MCP,
claude mcp list(names and status) is enough. To show a config's shape, print its keys and URLs, never a header value. If you print a secret by mistake, say so in the report's Notes for the operator: name the variable and the transcript path, and never the value.
Part 1 — the parts of VALIDATE-INSTALL.md that still apply from inside this notepad
Tasks 0–4 of that document (orient, identity, ground truth, hooks, preflight) are about the
KIT and do not depend on being in a notepad — if you need them, they are unchanged and live
at <kit-root>/vendor/dark-factory/starter-kit/instance/VALIDATE-INSTALL.md on a kit (the
vendorDir the lockfile names; starter-kit/instance/… only in a Tier-1 checkout). What
changes in a notepad session
is continuity, dispatch, and subagents, so those three are reproduced here adapted to it.
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.
- 21d ago Changed 174d78bd0bad
- 1mo ago Changed · +16 lines a4b44bf6be15
- 2mo ago Changed · +34 lines 29f55d06801c
- 2mo ago Changed · +22 lines 5eac948ca5dc
- 2mo ago First seen · 244 lines · 19 tokens per session scan C 6cdfe8c0bdc0
validate is a command published in the GitHub repository OneDro1d/dark-factory (3 stars, last pushed 4d ago), licensed Apache-2.0. It adds 19 tokens to every session and 5,358 once invoked, about $0.0001 per session on Opus 5.5. A static security scan graded it C with 1 finding (recursive force delete). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-09-09.
validate compared with commands from other repositories
| Mod | Type | Stars | Last push | Licence | Scan | Tokens | Works with |
|---|---|---|---|---|---|---|---|
| validateOneDro1d/dark-factory | Command | 3 | 4d ago | Apache-2.0 | C | 19 | Claude Code |
| checklistgithub/spec-kit | Command | 150k | today | MIT | A | 13 | Claude Code |
| clarifygithub/spec-kit | Command | 150k | today | MIT | A | 27 | Claude Code |
| specifygithub/spec-kit | Command | 150k | today | MIT | A | 13 | Claude Code |
| convergegithub/spec-kit | Command | 150k | today | MIT | A | 35 | Claude Code |
| implementgithub/spec-kit | Command | 150k | today | MIT | A | 14 | Claude Code |
| speckit.bug.fixgithub/spec-kit | Command | 150k | today | MIT | A | 12 | Claude Code |