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/innocarpe/CarpeOSnpx agentmods add skills/innocarpe/carpeos/carpeos-ciWrote 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/innocarpe/carpeos/carpeos-ci)<a href="https://agentmods.dev/skills/innocarpe/carpeos/carpeos-ci"><img src="https://agentmods.dev/badge/skills/innocarpe/carpeos/carpeos-ci/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/innocarpe/carpeos/carpeos-ci"><img src="https://agentmods.dev/badge/skills/innocarpe/carpeos/carpeos-ci.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 | $0.00106 | $0.02483 |
| Opus 5 | $0.00053 | $0.01241 |
| Sonnet 5 | $0.00021 | $0.00497 |
| Haiku 4.5 | $0.00011 | $0.00248 |
Grade A, and why
carpeos-ci 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 11d 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 — 231 lines — stays where its author put it; the contents beside it link to each section on GitHub.
CarpeOS CI (shared harness skill)
Same standard for Claude Code, Codex CLI, Grok Build, and Gajae Code/GJC.
Do not grow CI ad hoc. Load this skill and the policy SSOT before adding, moving, or “temporarily” wiring any check.
Source of truth
| Artifact | Path |
|---|---|
| Policy SSOT | docs/maintainers/ci-policy.md |
| Primary workflow | .github/workflows/ci.yml |
| Secret scan | .github/workflows/secret-scan.yml |
| Trust plane (gated) | .github/workflows/product-4-candidate-*.yml |
| Release | .github/workflows/release.yml |
| Local entrypoints | root package.json (check, test, smoke:*) |
| Agent rules | AGENTS.md / Agents.md (CI section) |
Read docs/maintainers/ci-policy.md in full before changing workflows or
check composition. This skill is the operating procedure; the doc is the contract.
When to use
Trigger on any of:
- edit under
.github/workflows/; - change
pnpm check/test/ smoke / e2e scripts used by CI; - “CI is slow”, “Actions is red noise”, “add a required check”;
- Product 4 evaluate / attest / publish workflow work;
- proposal to run heavy evals on every PR;
- after a merge that reintroduces invalid workflow YAML.
Hard rules (fail closed)
- Lanes first. Every check belongs to exactly one default lane:
pr-lean— merge-blocking, budgeted for iteration;main-full— deeper integration after merge;trust-release— Product 4 / release authority planes;local-only— agent/human, not GHA.
- PR lean budget. Target ≤ 2 minutes wall clock on
ubuntu-latestwith warm cache. If a change would push PR Checks > 3 minutes, stop and either re-home work tomain-full/ tests or get explicit maintainer approval in the PR Why section. - No duplicate monorepo work on PR. Do not rebuild packages already built by
pnpm check(or the lean equivalent) only to run a short eval on every PR. - Prefer tests over new workflows. New invariants land as package tests or
scripts/test/*.test.mjsfirst. New YAML needs the change-gate table. - Job-level
envmust not referencerunner.*. Use stepwith:/run:or a setup step +GITHUB_ENV. Invalid workflow files (0s, path-as-name) are P0. - Trust plane is not default PR CI. Product 4 candidate workflows are not required merge gates until activation policy says so. Logic is enforced by unit/contract tests; live GHA may be off, dispatch-only, or non-required.
- Release credentials never on PR. No npm tokens, signing keys, or
checks-write App privileges on untrusted
pull_requestjobs. - Same commands locally. Agents reproduce CI with root
pnpm/node --testscripts. Do not invent CI-only unverifiable glue. - Harness-neutral. Do not add “Claude-only” or “Grok-only” CI rules. Update
this skill +
docs/maintainers/ci-policy.mdinstead. - Honest PR validation. When CI changes, the PR validation table lists real
commands and results (
carpeos-prskill).
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.
- 11d ago First seen · 231 lines · 106 tokens per session scan A 89be89eb6602
carpeos-ci is a skill published in the GitHub repository innocarpe/CarpeOS (0 stars, last pushed 27d ago), licensed Apache-2.0. It adds 106 tokens to every session and 2,483 once invoked, about $0.0005 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 skills, from other repositories
wiki-build
A tool for building or rebuilding a local wiki from all source files in a vault, which is a folder used to store source material. It creates structured pages, layered indexes, logs, and cross-links.
knowledge-ingestion
A skill for saving links, social posts, videos, podcasts, screenshots, PDFs, local files, or text into an Obsidian knowledge base, a personal collection of searchable notes. The ingestion tools analyse and format the material for storage.
personal-knowledge-ingestion
A skill for saving links, videos, social posts, local files, screenshots, PDFs, or text into an Obsidian vault, a folder used for personal notes and knowledge. It uses connected knowledge tools to process and format the saved item.
remotion
Molio's builtin skill for MAKING a video from any source — wiki notes, articles, scripts, product info, or a brief — and rendering it to MP4. Use whenever the user wants to create/produce a video: a product demo, promo, trailer, intro, explainer, motion graphic, kinetic typography, social clip, or animated logo — even…
docling
PRIMARY skill for converting .pdf, .docx, .pptx, .xlsx, .doc, .ppt, .xls, images, and audio/video files (.mp3, .wav, .m4a, .mp4, .mov, etc.) to Markdown. Always prefer this over the docx/pdf/pptx/xlsx/paddleocr skills — docling uses GPU-accelerated OCR + layout detection + table structure extraction for documents, and…
wiki-ingest
A tool for adding source files or URLs to an existing local wiki. It creates or updates summary, topic, comparison, and index pages, with links between related pages.