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 cboone/agent-harness-plugins --skill review-branchgit clone --depth 1 https://github.com/cboone/agent-harness-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/cboone/agent-harness-plugins/review-branch)<a href="https://agentmods.dev/skills/cboone/agent-harness-plugins/review-branch"><img src="https://agentmods.dev/badge/skills/cboone/agent-harness-plugins/review-branch/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/cboone/agent-harness-plugins/review-branch"><img src="https://agentmods.dev/badge/skills/cboone/agent-harness-plugins/review-branch.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 | $0.04207 |
| Opus 5.5 | $0.00014 | $0.01683 |
| Sonnet 5.5 | $0.00007 | $0.00841 |
| Haiku 4.5 | $0.00003 | $0.00421 |
Grade A, and why
review-branch 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 yesterday.
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 — 380 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Review Branch
Review and evaluate all work done on the current branch: summarize changes, assess plan compliance, and evaluate code quality on its own merits.
Options
The user may provide these options inline:
- --plan
<path>: Path to a plan document to compare progress against - --since
<ref>: Use a specific tag, branch, or commit as the base reference instead of the default branch - --brief: Output only a high-level summary without the detailed breakdown
- --no-save: Skip saving the review to
docs/reviews/(default: save after outputting)
Capabilities and authority
Use the tools this session actually offers. For a question, use a structured question tool when available in the active mode; otherwise present numbered choices and wait. In a noninteractive run, report the unresolved question and preserved state instead of choosing an answer. Tool names such as Write, Read, Glob and AskUserQuestion below are examples: perform the same file or question operation with the host's supported mechanism. Write a body file completely before issuing the command that reads it; never batch those dependent operations.
Prior user authorization carries through the workflow; add no confirmation merely because the harness differs. Host approvals, active Plan or read-only mode, writable roots and network policy still apply. Stop before a prohibited write or any step dependent on a blocked operation, preserve state, and report the exact command and restriction. Never bypass controls or claim an unrun step completed. Separate completed, skipped, blocked and failed work in the result. For invocation hints, render the named skill with the running harness's form: slash invocation in Claude Code, dollar mention or the skills selector in Codex. A mention is not a tool call; load composed skills from their advertised locations.
Parent Continuation Contract
Honor a supplied parent continuation block and its resume target. On success, return control to the parent without asking whether to continue. When a decision has no supplied answer, do not ask independently or guess: return Status: needs-decision, the question, available choices and preserved state. The parent asks or stops, then invokes this skill with an answer. For an in-progress merge or rebase, list conflicts and the applicable abort command; do not abort or discard work merely to return a result.
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.
- yesterday Changed · +16 lines d5ddd714152a
- 28d ago Changed · +1 lines · -2 tokens per session 8ede0fba262f
- 1mo ago Changed · -5 lines · -69 tokens per session 79596d02d727
- 1mo ago Changed · +2 lines a5cd892142d5
- 2mo ago First seen · 366 lines · 105 tokens per session scan A 0e4cc82ac31f
review-branch is a skill published in the GitHub repository cboone/agent-harness-plugins (2 stars, last pushed today), licensed MIT. It adds 34 tokens to every session and 4,207 once invoked, about $0.0001 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-preflight
Review tooling for a PR: review-preflight prints REVIEW.md's criteria and the PR's live gate state before a review; ci-triage maps a PR's failing checks to already-filed issues.
pr-triage
4-phase PR backlog management with audit, deep code review, validated comments, and optional worktree setup. Use when triaging pull requests, catching up on pending code reviews, or managing a backlog of open PRs. Args: 'all' to review all, PR numbers to focus (e.g. '42 57'), 'en'/'fr' for language, no arg = audit…
pr
Analyze changes, detect scope issues, and create a well-structured PR.
validate-changes
Evaluate staged changes using LLM-as-a-Judge before committing.
review-pr
Comprehensive PR review -- full branch diff against base, commit-by-commit analysis, breaking change detection, conventional commit verification.
pr-audit
Audit a specified PR or PR-ready branch for submission or post-revision readiness. Bind the exact base/head, review commits and net changes, run safe project checks and inspect applicable CI. An audit alone does not authorize fixes, history changes or publication.