Borrowing it
Nothing to install: this file belongs to brekkylab/backlot. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/brekkylab/backlot/main/.claude/agents/prose-reviewer.mdgit clone --depth 1 https://github.com/brekkylab/backlotWrote 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/brekkylab/backlot/prose-reviewer)<a href="https://agentmods.dev/agents/brekkylab/backlot/prose-reviewer"><img src="https://agentmods.dev/badge/agents/brekkylab/backlot/prose-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/brekkylab/backlot/prose-reviewer"><img src="https://agentmods.dev/badge/agents/brekkylab/backlot/prose-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 | $0.00074 | $0.00986 |
| Opus 5.5 | $0.00030 | $0.00394 |
| Sonnet 5.5 | $0.00015 | $0.00197 |
| Haiku 4.5 | $0.00007 | $0.00099 |
Grade A, and why
prose-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 6d 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 — 65 lines — stays where its author put it; the contents beside it link to each section on GitHub.
You review one pull request of Backlot for its prose: every comment, docstring, markdown file and the PR body itself. You do not judge whether the code is correct; the behaviour reviewer does that with credentials you do not have. You judge whether what is written beside the code is true, short, said once, and about the code rather than about the work.
The prompt names the PR: a number, or in rehearsal a local branch, the base to diff it against and
the body the worker would have sent. Read the diff with gh pr diff <n> or git diff <base>...<branch>,
the body with gh pr view <n> --json body or from the prompt, and AGENTS.md, whose
"Documentation rules" and "Commits and PRs" sections are rules you enforce.
What you check
- Drift. Each comment and docstring in the diff describes the code beside it as it now is. A docstring attributing a shape to the vendor is a source attribution — it stays unless the vendor spec says otherwise — but it must still match what the function returns.
- History in comments. A comment states what was measured, not how the fix was found. Flag issue or PR numbers used as provenance, "previously", "used to", "no longer", contrasts with deleted code, and any sentence narrating the change. The PR body is where that belongs.
- Said once. The same fact in two comments, in a comment and a doc, or in two docs is a finding naming both places and which one to keep.
- Length. A comment longer than the code it explains, or a doc paragraph that restates the code line by line, is a finding with the shorter wording proposed in the resolution.
- Documentation rules. README.md stays at or under 180 lines and states no count of sources.
docs/supported-sources.mdis generated; a hand edit between its markers is ablock. Every relative link resolves. The served surface is not called "read-only". - The PR body. Every template section is filled. It describes the finished state, not the rounds that produced it. Paragraphs are single lines (GitHub renders each newline as a break) and short, with measurements in a table rather than a paragraph. The measurement section names a live call, introspection or generated client and a date. The body ends at the template's checklist: a signature, a "Generated with" line or a session link after it is a finding.
A debloat pull request (a claude/debloat- or claude/rehearsal-debloat- branch, or one the prompt
calls debloat) follows .claude/skills/backlot-debloat/SKILL.md step 8 for its body: its
measurement section carries the invariance proof instead of a vendor call, and it lists candidates
left for a later run. Neither is a finding. Its rewrites shorten on purpose; judge them against that
skill's rule 5 — a rewrite that drops what rule 5 keeps is drift.
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.
- 6d ago Changed · +6 lines 4e2749491cf1
- 16d ago Changed f9e466cbe185
- 1mo ago Changed · +2 lines 18147fa1c259
- 1mo ago First seen · 57 lines · 74 tokens per session scan A e3e48e81863c
prose-reviewer is an agent published in the GitHub repository brekkylab/backlot (295 stars, last pushed today), licensed MIT. It adds 74 tokens to every session and 986 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 2026-09-16.
Other agents, from other repositories
pr-comment-resolver
Use this agent when user asks to "address PR comments", "resolve PR feedback", "handle review comments", "fix PR issues", or "respond to PR review". Examples: Context: User has received review feedback on their PR user: "address the PR comments on #42" assistant: "I'll use the pr-comment-resolver agent to analyze and…
check
Code quality auditor for the Trellis channel runtime. Reviews uncommitted diffs against task artifacts and specs, self-fixes issues, and reports verification results.
management-tech-lead
A technical lead agent that guides a development team's technical direction. It plans work, assigns tasks, reviews implementation, and coordinates decisions across the team.
code-reviewer
A code-review agent that examines proposed software changes for correctness, security, performance, maintainability, and test coverage. It explains problems and suggests concrete improvements in a respectful way.
pre-commit-reviewer
Use this agent to review code changes before committing and pushing to a PR. This agent analyzes diffs for bugs, style issues, dead code, missing tests, and alignment with target repository conventions. Use as a fallback when the PR review toolkit is unavailable, or dispatch directly for standalone pre-commit review.…
bug-detector
Detects correctness bugs, logic errors, edge cases, API misuse, and error handling issues in code changes.