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/escoffier-labs/brigadenpx agentmods add skills/escoffier-labs/brigade/verify-brigadeWrote 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/skills/escoffier-labs/brigade/verify-brigade)<a href="https://agentmods.dev/skills/escoffier-labs/brigade/verify-brigade"><img src="https://agentmods.dev/badge/skills/escoffier-labs/brigade/verify-brigade/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/skills/escoffier-labs/brigade/verify-brigade"><img src="https://agentmods.dev/badge/skills/escoffier-labs/brigade/verify-brigade.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 | $0.00092 | $0.04269 |
| Opus 5 | $0.00046 | $0.02135 |
| Sonnet 5 | $0.00018 | $0.00854 |
| Haiku 4.5 | $0.00009 | $0.00427 |
Grade C, and why
verify-brigade 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 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.
Recursive force deletehighDestructive command
rm -rf with a variable or a broad path is one typo away from removing the wrong tree.
(`rm -rf .venv`). Do not `git add` either tree. How it starts
The opening of the file, as written. The whole thing — 354 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Verify Brigade
Brigade has no server and no UI. The surface a user touches is the brigade
CLI, and the only honest way to prove a change is to run that CLI against a
target and read what it wrote. This skill drives it through one helper,
control-brigade.py, so a proof is a rerunnable command instead of a story.
Every drive happens in a temp target this skill creates. The helper enforces
that: --target is accepted only when it resolves under <state-root>/targets,
so the operator home, the Brigade checkout, and every other path are refused
before a command runs. The state root gets the same treatment whether it comes
from --root or from the default - resolved, refused if it resolves to or
contains the home directory or the checkout, and refused if it is a symlink at
the root or at targets//captures/, so a link planted in a world-writable
temp dir cannot redirect a drive. cleanup removes only paths new-target
recorded. AGENTS.md forbids driving the real workspace anyway; here it is not
reachable.
What that paragraph does not claim:
- A path under the temp base is allowed even when
$TMPDIRsits inside the home directory. That is the only carve-out, and it is exactly that narrow: the home and checkout refusals run first, so the allowance applies only when the temp base itself resolves inside the home directory and outside the checkout. A$TMPDIRinside the checkout is still refused, and a relative$TMPDIRis refused outright rather than resolved against the working directory. The temp base itself is refused - a state root there would chmod the shared scratch tree to0700. - The default evidence root is
<checkout>/.brigade/verification-evidence, which is inside the checkout. That is the one deliberate exception: it is gitignored, and it sits outside the state root on purpose socleanupcannot reach the proof. The default and--evidence-rootare both guarded like--root, symlink check included: a link planted at either is refused, not followed.--labelnames the directory under that root, so it must be a simple[A-Za-z0-9._-]name of 1-64 characters - no separators, no... grokbot-feed --manifestis the one flag outside the target contract. It is a caller-supplied path, read for validation only and never written, so an operator can validate a real private feed manifest from wherever it lives. Everything the drive writes still lands in the target or the state root.
What ships with it
7 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
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 · 354 lines · 92 tokens per session scan C c4f4f6d94cd9
verify-brigade is a skill published in the GitHub repository escoffier-labs/brigade (72 stars, last pushed today), licensed MIT. It adds 92 tokens to every session and 4,269 once invoked, about $0.0005 per session on Opus 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 2026-09-05.
Other skills, from other repositories
human-pass
Guide real-build acceptance: literal taps, real inputs, expected outcomes, worst case first, one sitting. Triggers: testflight, human pass, acceptance test, before we ship, hand it to the user, what should I test, release checklist, device test.
playwright-testing
E2E testing with Playwright - Page Objects, cross-browser, CI/CD.
api-tester
A tool for creating and checking API tests from the real API contract and implementation. An API is the agreed way that software sends requests and receives responses.
visual-validation
Autonomous screenshot verification — captures, compares, and AI-evaluates UI changes for visual regressions.
memstack-development-webapp-testing
Use when the user says 'write browser tests', 'test this page', 'playwright test', 'e2e test', 'end to end test', 'browser test', 'test the UI', or needs Playwright-based browser testing for a web application. Do NOT use for unit tests, API tests, or non-browser testing.
qa-knowledge
To run QA engineering — requirements/gap analysis, scenario & spec design, test implementation, failure triage — over the QA knowledge base.