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 PrabhdeepSingh/claude-plugins --skill self-reviewgit clone --depth 1 https://github.com/PrabhdeepSingh/claude-pluginsWrote 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/prabhdeepsingh/claude-plugins/self-review)<a href="https://agentmods.dev/skills/prabhdeepsingh/claude-plugins/self-review"><img src="https://agentmods.dev/badge/skills/prabhdeepsingh/claude-plugins/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/prabhdeepsingh/claude-plugins/self-review"><img src="https://agentmods.dev/badge/skills/prabhdeepsingh/claude-plugins/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 | $0.00079 | $0.06085 |
| Opus 5.5 | $0.00032 | $0.02434 |
| Sonnet 5.5 | $0.00016 | $0.01217 |
| Haiku 4.5 | $0.00008 | $0.00609 |
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 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 — 158 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Self-review — point attention, don't bless the change
A self-review is not a score. It is not an approval. It is a pointer: here are the spots a reviewer should look hardest at, and here is why. Self-scoring rubber-stamps the model's own work — the model that wrote the code is the least reliable judge of whether it's right. That bias is structural, so on any substantial diff the reading is taken away from the author entirely: an independent reader on a cheaper model tier, which never saw this conversation, reads the diff cold against every checklist the diff's contents call for, and its findings then have to survive a default-reject bar in this session. Author-blindness is defeated by fresh eyes; finding-inflation is defeated by making every finding earn its place — and the read itself is priced by what the diff needs, not by how many checklists exist.
When to apply this
- Conductor hand-back (
/sonu:buildPhase 3) — before handing back to the user. - Ship pre-PR fix loop (
/sonu:shipPhase 1.5) — each pass of the loop runs this skill; the final pass's list seeds the## Risk / reviewer attentionsection of the PR body. - Standalone — any time the user says "self-review", "what should I look at", "what's risky here", or similar; or when a change is finished and about to be handed off. When invoked directly as
/sonu:self-review, review$ARGUMENTS— the text typed after the invocation; if that token appears literally or is empty, review the working tree / current branch per the diff-picker below.
How to apply this
1. Read the diff.
Pick the right diff command for the current state:
- Uncommitted changes (working tree dirty):
git diff HEAD— plusgit status --porcelainto catch untracked files. Brand-new files never added to the index do not appear ingit diff HEADat all; read each untracked file directly and include it in the review. A change made entirely of new files (a fresh module plus its tests) is the classic case wheregit diff HEADshows nothing while there is plenty to review. - Just committed (working tree clean, reviewing what was just committed):
git show HEADorgit diff HEAD^ HEAD - Staged only:
git diff --cached - Whole branch, about to open a PR (clean multi-commit branch):
git diff origin/<base>...HEAD(three dots = diff from the merge base). Reviewing only the last commit on a multi-commit branch misses risks in the earlier ones.
What ships with it
2 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.
- 1mo ago Changed · +4 lines 3522c622b717
- 1mo ago Changed · +2 lines · +5 tokens per session c2832f6a24b1
- 1mo ago Changed · +2 lines ecbeb07ddb7c
- 1mo ago First seen · 150 lines · 74 tokens per session scan A fa63ca7f12a4
self-review is a skill published in the GitHub repository PrabhdeepSingh/claude-plugins (4 stars, last pushed today), licensed MIT. It adds 79 tokens to every session and 6,085 once invoked, about $0.0003 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
review
Run a full code review on current changes or specified files. Runs ruff lint, review-code agent, and security scan in sequence.
java-jpa
Reviews Spring Data JPA for N+1 queries, fetch strategies, projections, and Specifications. Use when user asks to "review JPA", "check for N+1", "JPA performance", "review my entities", "check fetch strategy", or "review my repositories".
java-health
Runs a holistic code health check scoring Security, Tests, Performance and Quality with A-F grades. Use when user asks to "check health", "score this project", "health check", "how good is this code", "overall assessment", or "code quality score".
java-cache
Use when the user asks to add caching, configure Redis or Caffeine cache, use @Cacheable/@CacheEvict/@CachePut, optimize repeated database or API calls, or review existing Spring Boot cache configuration.
java-api-review
Reviews Java REST API design including HTTP methods, status codes, naming, and versioning. Use when user asks to "review my API", "check REST design", "is this good REST", "review endpoints", "API design review", "check my controller", or "review HTTP API".
java-concurrency-review
Reviews Java code for thread safety, race conditions, deadlocks, and Java 21 virtual thread compatibility. Use when user asks to "review concurrency", "is this thread safe", "check for race conditions", "concurrency issues", or "virtual thread compatible".