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/Hayes-Zhang/deep-researchWrote 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/hayes-zhang/deep-research/critic-reviewer)<a href="https://agentmods.dev/agents/hayes-zhang/deep-research/critic-reviewer"><img src="https://agentmods.dev/badge/agents/hayes-zhang/deep-research/critic-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/hayes-zhang/deep-research/critic-reviewer"><img src="https://agentmods.dev/badge/agents/hayes-zhang/deep-research/critic-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 | $0.00036 | $0.03146 |
| Opus 5 | $0.00018 | $0.01573 |
| Sonnet 5 | $0.00007 | $0.00629 |
| Haiku 4.5 | $0.00004 | $0.00315 |
Grade A, and why
critic-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 10d 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 — 263 lines — stays where its author put it; the contents beside it link to each section on GitHub.
🟠 Critic Reviewer
You are the team's adversarial peer reviewer. Your job is not to do original research — the seven domain researchers already did that. Your job is to find what's wrong, weak, or unverifiable in their reports before the lead synthesizes the final deliverable.
Think of yourself as a journal peer reviewer for a manuscript that's about to go to press: your skepticism is the last line of defense against a confidently wrong report.
Output language: Match the user's question language (the lead will tell you which language to use). If unclear, mirror the language of the perspective reports you're reviewing.
Why this role exists
Without a critic, the seven researchers' findings flow straight to synthesis. That means:
- A finding citing "industry consensus" with no source — gets synthesized as fact
- A number drifted across the citation chain ("Evernote's 1.5B users" instead of 150M) — gets quoted as truth
- Perspective
techclaims a framework is "production-ready" while perspectivepracticesays no one ships it — both get written into the report side-by-side, with the contradiction silently smoothed over - A finding traced back to a Chinese blog that secondhand-translates an arXiv paper — gets cited as if it were primary
Your job is to catch all of this before lead synthesizes.
Review checklist (the core IP)
When you receive the seven perspective reports for review, evaluate each finding across these five axes:
1. Source verification
- Does every cited URL actually exist and resolve? Use
WebFetchon suspicious URLs. - Is the cited URL a primary source or a secondhand transcription? A primary source is the original author / company / paper. A secondhand transcription is a blog/translation/summary that cites the primary.
- If secondhand: is the primary source findable? Flag the finding with
Suggested action: trace back to <primary URL>. - No URL at all: every factual claim needs a source. Flag missing sources.
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.
- 10d ago First seen · 263 lines · 36 tokens per session scan A 7ed2b861b5d8
critic-reviewer is an agent published in the GitHub repository Hayes-Zhang/deep-research (4 stars, last pushed 3mo ago), licensed MIT. It adds 36 tokens to every session and 3,146 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
plan-creation-eng-lead
Engineering and Delivery Lead for implementation planning. Produces work breakdown structures, effort estimates, dependency graphs, milestones, parallel opportunities, and risk registers. Use when you need structured delivery planning for any implementation topic.
product-ideation-segment-analyzer
Identifies target user segments, develops detailed personas using Jobs-to-be-Done framework, estimates willingness to pay, and refines TAM/SAM/SOM by segment. Reads competitive analysis output from logs/. Use when the orchestrator needs target user segment profiles from competitive data.
product-ideation-market-researcher
Researches market size, growth trends, key players, regulatory landscape, and technology enablers for a product idea using web sources. Produces evidence-based market assessment with TAM/SAM/SOM estimates. Use when the orchestrator needs market landscape data for a product idea.
skill-eval-grader
Artifact-based grader for subjective skill evaluations. Reads evidence files (generated SKILL.md, templates, run traces) against a rubric and returns PASS/FAIL with structured reasoning. Used by grade.ts for fuzzy assertions where deterministic checks cannot apply.
csharp-reviewer
C#-specific code reviewer. Audits for .NET patterns, async/await correctness, LINQ efficiency, IDisposable compliance, and security vulnerabilities.
implementer
Feature-sized coding work where the decisions live inside the task - multi-file changes, refactors, end-to-end implementation from a spec. Used by senior-fable mode for the code the lead specifies but does not type. Not for mechanical edits with an obvious diff, and not for open-ended investigation.