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 skills add aiatelie/ai-atelie --skill cuj-guardiangit clone --depth 1 https://github.com/aiatelie/ai-atelieWrote 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/aiatelie/ai-atelie/cuj-guardian)<a href="https://agentmods.dev/skills/aiatelie/ai-atelie/cuj-guardian"><img src="https://agentmods.dev/badge/skills/aiatelie/ai-atelie/cuj-guardian/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/aiatelie/ai-atelie/cuj-guardian"><img src="https://agentmods.dev/badge/skills/aiatelie/ai-atelie/cuj-guardian.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 | $0.00137 | $0.03197 |
| Opus 5.5 | $0.00055 | $0.01279 |
| Sonnet 5.5 | $0.00027 | $0.00639 |
| Haiku 4.5 | $0.00014 | $0.00320 |
Grade A, and why
cuj-guardian 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 1mo 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 — 183 lines — stays where its author put it; the contents beside it link to each section on GitHub.
cuj-guardian
A contributor workflow for AI Atelie. The CUJ ("Critical User Journey") is now a suite of focused journeys, not one big spec — the same contract, split into pieces so a slow agent doesn't kill the cleanup phase or block evidence on the PR.
This skill is dev-time only. It does not load into adapter sessions spawned by the editor.
What the suite proves
Four journeys under web/tests/e2e/journeys/. Together they prove the product still does its core job:
home-loads.spec.ts— home renders, the create-project form is visible + interactive.create-project.spec.ts— name + Create lands on/editorwith a validp_*id; project dir exists on disk; iframe canvas mounts.agent-edits-canvas.spec.ts— the load-bearing one. A real Claude Code agent (or whichever adapter is selected) takes a chat prompt, writes to disk, and the iframe paints the requested change.cleanup-snapshot.spec.ts— leak guard: any project namedJourney · *is force-deleted; suite ends with the project list it started with.
If any of these fails, the product is broken in a way users will notice. That's the contract; everything else (smoke.spec.ts, type-check, build) is cheap-pre-flight stuff.
Cleanup safety guarantee
The CUJ creates exactly one test project (under a fresh p_* id assigned by the New Project flow), works against it, and deletes it via DELETE /api/projects/<id> in a finally{} block — even on failure. The test enforces the safety property explicitly:
- Snapshot before: lists
web/projects/at the start, sorted. - Delta assertion at step 4: the captured
projectIdmust NOT appear in the pre-test snapshot. If it does, we somehow latched onto an existing project's id (serious bug); fail loud. - Cleanup at step 10: deletes by EXACT id, single-target API call, no wildcards.
- Snapshot after at step 11: post-test
web/projects/list MUST equal the pre-test list. If anything was leaked OR anything else disappeared, the test fails.
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.
- 1mo ago First seen · 183 lines · 137 tokens per session scan A bd49420abf9f
cuj-guardian is a skill published in the GitHub repository aiatelie/ai-atelie (11 stars, last pushed 4mo ago), licensed MIT. It adds 137 tokens to every session and 3,197 once invoked, about $0.0005 per session on Opus 5.5. 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 skills, from other repositories
pptx-html-fidelity-audit
Audit a python-pptx export against its source HTML deck, identify layout/content drift (footer overflow, cropped content, missing italic/em, lost styling, off-rhythm spacing), and re-export with strict footer-rail + cursor-flow layout discipline. Use this skill whenever the user has a .pptx that was generated from an…
build-test
Run the project's build / typecheck / lint / test commands and emit the build.passing + tests.passing signals devloop convergence reads.
web-app-qa
Test a web app the way a real user would: click through the main flows, try odd inputs, watch for errors, and keep screenshots as proof. You get a bug report ranked by how serious each problem is.
release-parity-test
Use before cutting a release, after a change to live reads, write tools or skills, or when the user asks to test the MCP against the web app. Runs the three-way parity test in test-matrix.md (cache mode vs live MCP vs app.copilot.money), including round-trip writes with throwaway test data in the user's Manual…
nexu-e2e-test
Use when verifying OpenClaw gateway fixes end-to-end, testing skill loading after restart, or running integration tests against the local Nexu+OpenClaw stack. Triggers on "e2e test", "verify fix", "test gateway", "test skills loading".
mandu-testing
Testing patterns for Mandu applications. Use when writing unit tests, integration tests, or E2E tests. Triggers on test, spec, Bun test, Playwright, or testing tasks.