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 joe-broadhead/open-cowork --skill autoresearchgit clone --depth 1 https://github.com/joe-broadhead/open-coworkWrote 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/joe-broadhead/open-cowork/autoresearch)<a href="https://agentmods.dev/skills/joe-broadhead/open-cowork/autoresearch"><img src="https://agentmods.dev/badge/skills/joe-broadhead/open-cowork/autoresearch/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/joe-broadhead/open-cowork/autoresearch"><img src="https://agentmods.dev/badge/skills/joe-broadhead/open-cowork/autoresearch.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector pass
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 | $0.00000 | $0.02232 |
| Opus 5 | $0.00000 | $0.01116 |
| Sonnet 5 | $0.00000 | $0.00446 |
| Haiku 4.5 | $0.00000 | $0.00223 |
Grade A, and why
autoresearch 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 11d 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 — 284 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Autoresearch
Use this skill when the user wants autonomous, measurable iteration: "run auto research", "run autoresearch", "optimize this skill", "improve this agent", "improve this prompt with evals", "make this benchmark better", or "iterate until the metric improves".
The core pattern is Karpathy's autoresearch ratchet:
- Constrain the target and editable scope.
- Establish a baseline.
- Change one thing.
- Run mechanical verification.
- Keep only improvements; discard regressions.
- Log every experiment.
- Repeat until the budget, plateau rule, or user stop condition is reached.
Open Cowork owns composition only. Do not build a separate runner. Use OpenCode-native editing, shell, approvals, questions, tasks, and skill loading.
Setup Gate
Do not start experiments until these fields are known:
- Goal: what should improve.
- Target: file path, installed custom skill name, prompt, benchmark, or repo area.
- Mutable scope: exact files or bundle content that may change.
- Read-only scope: files, tests, evaluator code, fixtures, package files, or packaged builtin skills that must not change.
- Metric: a parseable score or binary eval suite.
- Direction: higher is better, lower is better, or pass count is better.
- Verify command or eval protocol: how each candidate is measured.
- Budget: iteration count, time cap, or explicit unbounded run.
- Apply policy: whether to write only a candidate, or apply the final improvement after approval.
If the user did not specify a budget, suggest 10 iterations. Only run unbounded or overnight loops when the user explicitly asks for that.
Target Modes
Custom Skill Optimization
When optimizing an installed Open Cowork custom skill:
- Use the Skills MCP to list and read the custom bundle when available.
- Copy
SKILL.mdand supporting files into the run directory. - Mutate the working copy during experiments.
- Do not overwrite the installed custom skill during the loop.
- At the end, ask for or rely on the approval prompt for
skills_save_skill_bundlebefore applying the final version.
What ships with it
1 file beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
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.
- 11d ago First seen · 284 lines · 0 tokens per session scan A fe1309c26aa7
autoresearch is a skill published in the GitHub repository joe-broadhead/open-cowork (11 stars, last pushed 6d ago), licensed MIT. It costs nothing until one of its globs matches a file; then it loads 2,232 tokens. 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
research-orchestrator
Becomes a principal research scientist who orchestrates deep, multi-source research by coordinating evidence gathering, grading source quality, and synthesizing findings with explicit confidence levels. Use when the user needs a comprehensive research report on a complex topic, evidence-based analysis with source…
peer-review
Structured manuscript/grant review with checklist-based evaluation. Use when writing formal peer reviews with specific criteria methodology assessment, statistical validity, reporting standards compliance (CONSORT/STROBE), and constructive feedback. Best for actual review writing, manuscript revision. For evaluating…
scientific-critical-thinking
Evaluate scientific claims and evidence quality. Use for assessing experimental design validity, identifying biases and confounders, applying evidence grading frameworks (GRADE, Cochrane Risk of Bias), or teaching critical analysis. Best for understanding evidence quality, identifying flaws. For formal peer review…
scientific-schematics
Create publication-quality scientific diagrams using Nano Banana 2 AI with smart iterative refinement. Uses Gemini 3.1 Pro Preview for quality review. Only regenerates if quality is below threshold for your document type. Specialized in neural network architectures, system diagrams, flowcharts, biological pathways…
scientific-brainstorming
Creative research ideation and exploration. Use for open-ended brainstorming sessions, exploring interdisciplinary connections, challenging assumptions, or identifying research gaps. Best for early-stage research planning when you do not have specific observations yet. For formulating testable hypotheses from data use…
imap-smtp-email
Read and send email via IMAP/SMTP. Check for new/unread messages, fetch content, search mailboxes, mark as read/unread, and send emails with attachments. Works with any IMAP/SMTP server including Gmail, Outlook, 163.com, vip.163.com, 126.com, vip.126.com, 188.com, and vip.188.com.