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 agentmods add skills/bengous/claude-code-plugins/critiquenpx skills add bengous/claude-code-plugins --skill critiquegit clone --depth 1 https://github.com/bengous/claude-code-pluginsWhat 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 | $0.00075 | $0.01176 |
| Opus 5 | $0.00037 | $0.00588 |
| Sonnet 5 | $0.00015 | $0.00235 |
| Haiku 4.5 | $0.00007 | $0.00118 |
Grade A, and why
critique 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 2d 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 — 87 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Cross-model critique
A second pair of eyes from a non-Claude model (Codex) on a proposal Claude just made — a design, refactor, API, or fix. Goal: a genuine cross-model check. Validate what holds up, challenge what is weak, surface a better path when one exists. Not a rubber stamp, not reflexive contrarianism.
When to use
- Claude (this instance or another) proposed a solution and you want it pressure-tested before committing.
- You said "not bad, right?" and actually want the honest answer.
- A decision has real forks and one outside viewpoint would de-risk it.
NOT for reviewing committed code against a diff — use /codex:review or
/codex:adversarial-review. Those read the git diff; a proposal usually
lives in the conversation, not on disk, so a diff-based review would miss it
(and may review unrelated working-tree files instead).
Workflow
-
Capture the proposal — grounded. Write to one temp file: the proposal verbatim, the problem it solves, any constraints, the questions you most want challenged, an explicit list of the real repo file paths it touches or depends on, and — if the user gave extra focus with the invocation — a final
## Extra focus from the usersection carrying it verbatim. Grounding in actual code is the one thing that makes the critique useful; skip it and the review drifts into generic advice. Everything user-authored goes in the file, never inline in the shell command (quotes/backticks break inline prompts).Write /tmp/critique-proposal.md -
Run Codex read-only (it is a review; it must not edit). The substance, including any user focus, is in the file; the inline prompt only carries the proposal-file path — keep it in sync if you relocate the temp file. Capture JSONL so the thread id can be read back for follow-ups:
"${CLAUDE_PLUGIN_ROOT}/scripts/codex" exec \ -s read-only \ -c model_reasoning_effort=xhigh \ --json -o /tmp/critique-verdict.md \ "You are giving a CRITICAL SECOND OPINION on a proposal made by another AI (Claude), at the user's request. Read /tmp/critique-proposal.md in full — including any 'Extra focus from the user' section — then read the real repo files it lists before judging. Then: (1) briefly confirm what is sound; (2) challenge only what is genuinely weak — correctness bugs, wrong assumptions, missed edge cases, or a simpler/safer/more idiomatic option — grounding every point in the actual code; (3) where a better path exists, describe it concretely and explain WHY (tradeoffs); (4) if it is good as-is, say so plainly and do not invent problems. End with a one-line verdict: SHIP / ADJUST / RECONSIDER." \ </dev/null > /tmp/critique.jsonl
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.
- 2d ago First seen · 87 lines · 75 tokens per session scan A 737f45d725fb
critique is a skill published in the GitHub repository bengous/claude-code-plugins (4 stars, last pushed 2d ago), licensed MIT. It adds 75 tokens to every session and 1,176 once invoked, about $0.0004 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
sql-reporting
Conventions and review steps for writing analytics SQL against the warehouse. Use whenever the task involves querying tables, building a report, or aggregating metrics.
html-ppt-hermes-cyber-terminal
OpenDesign + BYOK: choosing and wiring your own model, hands-on — cost, quality, and the routing decision. Built as a decision-grade AI literacy deck for engineers, IT, applied-AI teams.
verify-security
安全校验关卡。自动扫描代码安全漏洞，检测危险模式，确保安全决策有文档记录。当用户提到安全扫描、漏洞检测、安全审计、代码安全、OWASP、注入检测、敏感信息泄露时使用。在新建模块、安全相关变更、攻防任务、重构完成时自动触发。.
development
开发语言能力索引。Python、Go、Rust、TypeScript、Java、C++、Shell。当用户提到编程、开发、代码、语言时路由到此。.
post-build-flow
Handles workflow verification and setup after build-workflow succeeds, or when the message contains workflow-verification-follow-up or workflow-setup-required. Load after direct builds, when verificationReadiness requires action, or on orchestrator verify/setup follow-up turns.
n8n:human-like-code-review
Reviews a GitHub pull request like a thoughtful human reviewer and writes the feedback to a markdown file. Prioritizes context, architecture fit, solution complexity, bugs, security edge cases, and missing tests. Use when given a PR URL to review, or when the user says /human-like-code-review.