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 firatcand/forge --skill auto-reviewgit clone --depth 1 https://github.com/firatcand/forgeWrote 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/firatcand/forge/auto-review)<a href="https://agentmods.dev/skills/firatcand/forge/auto-review"><img src="https://agentmods.dev/badge/skills/firatcand/forge/auto-review.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 | $0.00059 | $0.02231 |
| Opus 5 | $0.00030 | $0.01115 |
| Sonnet 5 | $0.00012 | $0.00446 |
| Haiku 4.5 | $0.00006 | $0.00223 |
Grade A, and why
auto-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 8d 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 — 202 lines — stays where its author put it; the contents beside it link to each section on GitHub.
/auto-review
Drain the orchestrator review-queue: for each task sitting in
ready_for_review, run the review, compose a single verdict, and either advance
the task to reviewed or escalate an architectural finding to you for a
decision.
This is a per-turn recipe — NOT a loop
This skill performs one drain pass over the head of the queue per
invocation: it reviews exactly one task, acts on it, and returns. It does
not sleep, poll, watch, or while-loop, and it never runs in headless mode.
The repeat-until-empty behavior is owned by the native /goal driver — you
run /goal "the review-queue is empty", and the host re-invokes this recipe
each turn until forge orchestrate review-queue --json reports an empty queue.
Keeping the loop in /goal (not here) is what lets every round run as a fresh,
fully-budgeted interactive turn.
Interactive session only — billing invariant
This skill runs only in an interactive session, for two reasons:
- Escalation needs structured input. An architectural finding is surfaced
to you through the host's
AskUserQuestion/Tasktooling — there is no non-interactive path for a human decision. - Billing invariant. The primary review runs as an in-session subagent
under your subscription via the host
Tasktool. For the primary review this skill NEVER shells out toclaude -p,claude --print, the Anthropic API, or any headless metered invocation. The only external reviewer it may call is the configuredsecond-opinionhost (Codex, Gemini, or Claude), and only on a critical path. Note: asecond-opinionhost ofclauderuns via meteredclaude -p(FORGE-224) — that is the second-opinion path, distinct from the subscription-billed in-session primary review above.
Skill ↔ verb contract
This skill owns the UX (queue read, review orchestration, the compose decision,
confirmations). Every state change goes through a CLI verb:
forge orchestrate complete (advance) or forge orchestrate question
(escalate/park). The skill NEVER mutates orchestrator state by hand and NEVER
writes into .forge/orchestrator/ directly.
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.
- 8d ago First seen · 202 lines · 59 tokens per session scan A 022a5ea3b72d
auto-review is a skill published in the GitHub repository firatcand/forge (13 stars, last pushed 1mo ago), licensed MIT. It adds 59 tokens to every session and 2,231 once invoked, about $0.0003 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 2026-08-30.
Other skills, from other repositories
prospec-verify
A review skill that checks an implementation against its tasks, written requirements, project rules, supporting knowledge, tests, and—when relevant—its design.
prospec-review
Adversarial Code Review → Fix Loop - Between implement and verify, an independent fresh-context reviewer audits the whole change diff; verifier-confirmed criticals are auto-fixed, majors are proposed, and a spec-aware lens checks delta-spec/dependency-direction. Triggers: review, code review, adversarial review, find…
ast-grep
Code search, analysis, and refactoring using ast-grep (sg). Use for AST-based code modifications, structural search, and linting.
cleanup-codebase
Reduce concepts, duplication, and ceremony in internal code while touching nearby code. Use when working an existing path and you spot dead fields, redundant wrappers, or speculative abstractions; distinct from refactor-break-bw-compat (internal hygiene, not public API removal).
pr-reviews
Review code changes on a given GitHub PR using gh CLI. Use when the user asks to review a pull request, analyze PR diffs, or provide feedback on open PRs with structured quality, security, and testing assessments.
reviews
Review the code changes on the current branch. Use when the user asks to review their current work, analyze recent commits, or get a code quality assessment of the active branch against the main branch.