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 corveil/crow --skill crow-release-reviewgit clone --depth 1 https://github.com/corveil/crowWrote 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/corveil/crow/crow-release-review)<a href="https://agentmods.dev/skills/corveil/crow/crow-release-review"><img src="https://agentmods.dev/badge/skills/corveil/crow/crow-release-review/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/corveil/crow/crow-release-review"><img src="https://agentmods.dev/badge/skills/corveil/crow/crow-release-review.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 | $0.00139 | $0.09747 |
| Opus 5.5 | $0.00056 | $0.03899 |
| Sonnet 5.5 | $0.00028 | $0.01949 |
| Haiku 4.5 | $0.00014 | $0.00975 |
Grade C, and why
crow-release-review scanned grade C with 2 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 today.
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.
1. Clear the last read, so a doc deleted since then doesn't linger: `rm -rf D/raw`. Makes network callslowCapability
Not a fault in itself. Listed so you know the mod talks to something, and to what.
**Every agent step stays read-only against prod**: describe, list and get calls, CLI reads, curl GETs, log queries. A step that would write belongs to a person (`human`). How it starts
The opening of the file, as written. The whole thing — 428 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Crow Release Review Skill
Purpose
Packages the release a Manager prepared and shipped by hand on 25 Oct 2026: three repos, 45 PRs, one page. The operator asked what was on main since each repo's last release. For each change they wanted what changes for users, a plain-language before and after with pros and cons, how risky it is to deploy, release blockers, post-release test steps a coding agent can run, and a notes box. Then they ran the whole release from that page.
- Study what is on
mainsince each repo's last release, and how each repo releases. - Publish one release review page: a summary, one tab per repo, a runbook, and a test ledger.
- Read back the operator's notes, decisions and runbook ticks.
- Run the release with them: baselines before each tag, and each tag only on their explicit go.
- Verify after each release goes live: run the agent-runnable steps and record pass or fail with evidence on the page.
In scope: the study, the page, its storage contract, the read-back, and the verify loop. Out of scope: tagging or deploying on your own. You run a tag command only when the operator says go for that tag, because every tag is a prod deploy. Fixing what the study finds is out of scope too: blockers and findings become tickets through /crow-create-ticket.
Requirements
- A Manager running Claude Code signed in to claude.ai, with the
ArtifactandArtifactDatatools. git, andghauthenticated for every repo.jq1.6 or newer.- A local clone of each repo (see Phase 0).
- Whatever read-only access the test steps need (a cloud CLI profile, a product CLI). The study names it per step in
needs.
Without the Artifact tools (Cursor, Codex, OpenCode Managers, or Claude Code without a claude.ai login), the page can't be published. Say so, then fall back to a Markdown report:
R report --items D/items.json --out D/report.md
Show the verdict, the blockers and the runbook in chat, give the path to the full report, and ask the operator for decisions and notes in chat. The rest of the loop still works on disk: record each step result yourself as D/raw/checks/{step id}.json with the same shape as the page (The storage contract), and R report --items D/items.json --data D/raw shows them as ticked or failed.
What ships with it
3 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.
- today First seen · 428 lines · 139 tokens per session scan C a625c15129ba
crow-release-review is a skill published in the GitHub repository corveil/crow (20 stars, last pushed today), licensed Apache-2.0. It adds 139 tokens to every session and 9,747 once invoked, about $0.0006 per session on Opus 5.5. A static security scan graded it C with 2 findings (recursive force delete, makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-10-28.
Other skills, from other repositories
batch
Research and plan a large-scale change, then execute it in parallel across 5–30 isolated worktree agents that each open a PR.
team-release
Orchestrate the release team — release-manager, qa-lead, devops-engineer, producer — to execute a release from candidate to deployment.
operating-github-cli
Default GitHub skill for the action agent. Use githubcli for any GitHub request — create/list/view/close issues and PRs, assign, labels, repos, releases, checks, github.com/owner/repo URLs, or gh api. Prefer over shellrun/!gh. Run these with githubcli in the current agent turn.
github
Drive GitHub via the official gh CLI — repos, issues, pull requests, releases, gists, Actions runs, and raw REST through gh api. Use when the user asks to inspect or manage GitHub.
testdino-releases
Use when the user wants to browse, inspect, create, or update releases/milestones in a TestDino project. Covers listreleases, getrelease, createrelease, and updaterelease. Accepts counter-style IDs like MS-12.
create-milestone
Create a GitHub milestone for an upcoming release. Suggests the next version based on the latest release, gathers all merged PRs and closed issues since that release, presents a draft with two tables (Issues and PRs) for user approval, then creates the milestone and assigns all approved items.