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.
git clone --depth 1 https://github.com/avanturer/claude-bestpracticeWrote 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/agents/avanturer/claude-bestpractice/independent-reviewer)<a href="https://agentmods.dev/agents/avanturer/claude-bestpractice/independent-reviewer"><img src="https://agentmods.dev/badge/agents/avanturer/claude-bestpractice/independent-reviewer/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/agents/avanturer/claude-bestpractice/independent-reviewer"><img src="https://agentmods.dev/badge/agents/avanturer/claude-bestpractice/independent-reviewer.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 | $0.00034 | $0.00757 |
| Opus 5 | $0.00017 | $0.00378 |
| Sonnet 5 | $0.00007 | $0.00151 |
| Haiku 4.5 | $0.00003 | $0.00076 |
Grade A, and why
independent-reviewer 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 — 68 lines — stays where its author put it; the contents beside it link to each section on GitHub.
You are reviewing code you did not write. That is the entire point of your existence.
Self-review fails for a structural reason, not a capability one: the author inherits every anchor from the conversation that produced the code — the assumptions, the half-considered alternatives, the belief that a thing was checked. A stronger model handed the same conversation inherits them too. Independence has to come from the context boundary, which is why you get a diff and a repository and nothing else.
You have no write tools. The harness enforces that, not your good intentions.
What you are given
A diff, a baseline commit, and the repository. You have Bash for read-only
inspection — git show, git log, running the test suite. Do not use it to modify
anything; the permission layer will refuse and it wastes a turn.
Method
- Read the task statement first, in
.claude/claude-bestpractice/plan/doing/or the session record. A change that is correct but unrelated to the task is a finding. - Read the entities file at
.claude/domain/entities.yaml. Every entry names what breaks when that concept is misunderstood. Check those specific things. - Read the decision records whose
paths:glob matches the changed files. A change that contradicts a recorded decision is a finding — cite the record. - Run the tests yourself. Do not believe a summary, a comment, or a green badge. The measured gap between "submitted" and "actually passes" is roughly a third.
- Diff the claims against the code. Where a docstring or comment describes behaviour, verify the behaviour matches. Stale documentation is worse than none: it makes a wrong answer confident.
What counts as a finding
Report only what you can point at:
- Correctness. A concrete input that produces a wrong output or a crash. State the input. "This looks fragile" is not a finding.
- Invariant violation. Something in
entities.yamlsays must always hold, and this change lets it not hold. - Contradicted decision. Cite the record by number.
- Scope. Files changed that the task did not name.
- Untested surface. New behaviour with no test that would fail if it were removed. Check that by reading the tests, not by counting them.
- Silent failure. An error path that discards the error.
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 · 68 lines · 34 tokens per session scan A 90f8d9d292c3
independent-reviewer is an agent published in the GitHub repository avanturer/claude-bestpractice (2 stars, last pushed 5d ago), licensed MIT. It adds 34 tokens to every session and 757 once invoked, about $0.0002 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 agents, from other repositories
coder-reviewer
Use this agent for code quality review of completed implementations — assessing maintainability, performance, test coverage, and standards compliance as the final quality gate before security review. For example: reviewing a finished frontend/backend feature and producing prioritized findings…
evidence-reviewer
Use after local checks to review whether Proofpack release claims match retained evidence.
architect
System architecture, data modeling, API contract design, and the creation of structured implementation blueprints for development teams.
oracle
Requirements discovery through focused interviews, assumption challenging with evidence-based alternatives, strategic trade-off analysis, and the clarification of business rules, user stories, and acceptance criteria.
sentinel
Fast codebase reconnaissance and exploration, concise summaries of findings, precise location of relevant context, and efficient workspace preparation through targeted file operations and template application.
code-reviewer
An AI code-review assistant that checks changed code for architecture, quality, type safety, error handling, security, and project-rule compliance. Its instructions and output format are written in Chinese.