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 swDomass/AI_orchestrator --skill review-loopgit clone --depth 1 https://github.com/swDomass/AI_orchestratorWrote 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/swdomass/ai_orchestrator/review-loop)<a href="https://agentmods.dev/skills/swdomass/ai_orchestrator/review-loop"><img src="https://agentmods.dev/badge/skills/swdomass/ai_orchestrator/review-loop.svg" alt="Measured on agentmods" 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 | $0.00023 | $0.00545 |
| Opus 5 | $0.00012 | $0.00272 |
| Sonnet 5 | $0.00005 | $0.00109 |
| Haiku 4.5 | $0.00002 | $0.00055 |
Grade A, and why
review-loop 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 4d 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.
What it actually says
System Prompt Addition
Perform an iterative code review. Classify findings as:
- P1 (blocker): bugs, security issues, data loss risks — blocking
- P2 (important): performance problems, maintainability issues — blocking
- P3 (minor): style, naming, minor improvements — NOT blocking
After each round, fix all P1 and P2 findings, then re-review. The loop ends when no
P1 or P2 remains — not when the finding list is empty. Max 20 iterations
(config.TOOL_MAX_ITERATIONS).
P3 is not fixed here. The tool removes every P3 from the fix prompt, so no P3 is ever requested; they are collected across all iterations and appended once to the final output as an offer, with file:line — the user decides. Do not ask for P3 fixes in this loop.
Scope of that guarantee: it covers the prompt, not the model's whole context. With
CLAUDE_SESSION_ENABLED=true the review and fix calls share one conversation
(SessionContext, --resume), so a P3 named in an earlier review is still in history and
the fixing step can see it. What holds in both modes is the part that matters: no P3
reaches the fix prompt, and — because the success gate counts only blocking findings — a
P3 can never keep the loop running. A fresh session before every write call would close
the gap, at the cost of the prompt-cache benefit session reuse exists for; that trade was
declined deliberately.
Why: touching working code for cosmetics widens the diff without functional gain, and because the re-review re-reads the diff fresh, every P3 fix produces new diff that can surface new P3 — the loop feeds itself and burns iterations on style. Fixing P1/P2 only is the direct application of "minimal impact", not a shortcut.
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.
- 4d ago Changed · +22 lines · +4 tokens per session d8a19a2fbba3
- 8d ago First seen · 24 lines · 19 tokens per session scan A 09484c75ee7a
review-loop is a skill published in the GitHub repository swDomass/AI_orchestrator (5 stars, last pushed 2d ago), licensed MIT. It adds 23 tokens to every session and 545 once invoked, about $0.0001 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
code-review-expert
Expert-level code review focusing on quality, security, performance, and maintainability. Use this skill for conducting thorough code reviews, identifying issues, and providing constructive feedback.
code-review
Automated code review with checklist and quality gates.
autonomous
Use when five specialized autonomous agents (code, deploy, planning, research, review) working as a coordinated pipeline. From spec to shipped code with automated planning, research, review, and deployment gates. Use when working with autonomous agents.
code-review
A code-review and quality-analysis skill written in Chinese. It is intended to inspect code, find potential problems, and suggest improvements.
code-reviewer
A code-review workflow written in Chinese that checks software changes for correctness, security, performance, and maintainability. It also defines a report format for critical issues and suggestions.
code-review
Code review for issues, style consistency, best practices—use for full code evaluation; not for debugging specific errors or fixing individual style violations.