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 hamr0/agentic-toolkit --skill self-reviewgit clone --depth 1 https://github.com/hamr0/agentic-toolkitWrote 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/hamr0/agentic-toolkit/self-review)<a href="https://agentmods.dev/skills/hamr0/agentic-toolkit/self-review"><img src="https://agentmods.dev/badge/skills/hamr0/agentic-toolkit/self-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/hamr0/agentic-toolkit/self-review"><img src="https://agentmods.dev/badge/skills/hamr0/agentic-toolkit/self-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 | $0.00020 | $0.02268 |
| Opus 5.5 | $0.00008 | $0.00907 |
| Sonnet 5.5 | $0.00004 | $0.00454 |
| Haiku 4.5 | $0.00002 | $0.00227 |
Grade A, and why
self-review 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 14d 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.
This is a copy
100% identical to self-review — 0 lines differ, which has more behind it and is treated as the original. This page carries a canonical link to it rather than competing with it.
How it starts
The opening of the file, as written. The whole thing — 149 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Answer "verify what you delivered, what did you gloss over, what did I miss?"
about committed work since the last self-review, before /branch-review. The
orchestrator resolves the range and hands off; a spawned worker tries to
break it.
Guardrails
- Spawn a worker, mid tier stated explicitly (omitted inherits the parent's, not the balanced one; not cheapest/fastest either — judgment degrades there; never a vendor model name — this spec names the tier, you translate it to whatever your spawn tool calls its mid tier), and hand it this file's path — a worker has no skill text of its own. No delegation, no sub-spawning — every command the worker cites is one it ran itself.
- Ask and surface only. Never fixes anything. The user picks.
1. Orchestrator — resolve the range, then hand off
Dirty tree first: git status --porcelain prints any line → stop: the
tree is dirty (list the paths), /self-review reviews commits not the working
tree, commit then re-run. No worker spawned.
Hashes or a range given ($ARGUMENTS: one or more hashes, or <a>..<b>)
→ review exactly those commits, nothing else, on any branch: each hash via
git show <sha>, a range via git log <a>..<b> and git diff <a>..<b>.
Validate each hash, and both ends of a range, with
git rev-parse --verify <x>^{commit}; reject anything starting with -.
Hash/range mode never rewrites self-review-sha:.
Neither → on main/master, stop and ask for hashes or a range.
Otherwise the bookmark is one line, self-review-sha:, in /branch-review's
record (/self-review is its only writer):
grep '^self-review-sha:' .amp/remember/last-review.md || grep '^debrief-sha:' .amp/remember/last-review.md
No self-review-sha: line → the old debrief-sha: name, once (that is what
the || does; the new name wins if both exist). Fails either check, starts
with -, or the third exits 0 (an ancestor of main's merge-base is from an
already-merged branch) → no bookmark:
git rev-parse --verify <sha>
git merge-base --is-ancestor <sha> HEAD
git merge-base --is-ancestor <sha> $(git merge-base main HEAD)
Valid (third command exits non-zero) → range <sha>..HEAD. No bookmark →
whole branch, git merge-base main HEAD..HEAD. Range empty → "nothing new
since the last self-review," stop, no worker spawned. Otherwise hand off —
what was done, claims made (works / tested / done), files changed, the
baseline suite totals if known, loose ends only you can know (a peer session
not told, a silent open question, unshipped state) — and spawn one mid-tier
worker.
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.
- 14d ago Changed · +35 lines 4214f17a0739
- 15d ago First seen · 114 lines · 20 tokens per session scan A 7a43fdd59cef
self-review is a skill published in the GitHub repository hamr0/agentic-toolkit (22 stars, last pushed 11d ago), licensed Apache-2.0. It adds 20 tokens to every session and 2,268 once invoked, about $0.0001 per session on Opus 5.5. A static security scan graded it A with 0 findings. It is 100% identical to self-review, differing in 0 lines, and is treated as a copy.
Other skills, from other repositories
karpathy-guidelines
A set of coding guidelines based on observations about common mistakes made by language models. It emphasizes simple solutions, small targeted edits, clear assumptions, and checkable results.
superpowers-receiving-code-review
Use when receiving code review feedback, before implementing suggestions, especially if feedback seems unclear or technically questionable - requires technical rigor and verification, not performative agreement or blind implementation.
github
A guide for using GitHub through the gh command-line tool, which lets developers work with GitHub without opening its website. It covers issues, pull requests, continuous-integration runs, code reviews, and API queries.
superpowers-requesting-code-review
Use when completing tasks, implementing major features, or before merging to verify work meets requirements.
coding-agent
A tool that sends coding work to an external coding agent already installed on the computer, such as Claude Code, OpenCode, or Pi. It is intended for feature development, pull-request reviews, large refactors, and exploratory coding.
requesting-code-review
A procedure for asking another coding agent to inspect completed changes before they are merged. Code review means checking whether the implementation meets its requirements and finding defects.