ClawHub

Openclaw Autoresearch

v0.2.4

Use when user wants to optimize, improve, benchmark, or evaluate a skill's prompt. Triggers on "optimize skill", "improve skill prompt", "benchmark skill", "...

⭐ 1· 117·0 current·0 all-time
byZHANG Ning@zning1994
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
VirusTotalVirusTotal
Benign
View report →
OpenClawOpenClaw
Benign
high confidence
✓
Purpose & Capability
Name/description (autoresearch / prompt optimization) match the included code and instructions. Required binaries (python3) and the primary env (OPENAI_API_KEY) are expected for a Python script that calls OpenAI-compatible APIs. The script implements MiniMax, OpenAI, and Anthropic providers as declared.
ℹ
Instruction Scope
SKILL.md and autoresearch.py clearly instruct the agent to read a target SKILL.md (or similar prompt file), run experiments, and send test inputs and prompt text to an LLM provider for mutation/judging. This is consistent with the stated purpose, but it necessarily transmits the target prompt and test inputs to remote LLM endpoints (or whatever OPENAI_BASE_URL you set). Ensure target files do not contain secrets or credentials you don't want sent to a third party.
✓
Install Mechanism
No install spec — the project is instruction+script only. There are no third-party downloads, no archive extraction, and the Python script uses only the standard library. Running the script will create output files (results.tsv, results.json, dashboard.html, SKILL.md.baseline) in the working directory as documented.
✓
Credentials
Only LLM API keys are requested (MINIMAX_API_KEY, OPENAI_API_KEY, ANTHROPIC_API_KEY) and an optional OPENAI_BASE_URL; these map directly to the declared provider integrations. No unrelated credentials or config paths are required. Note: OPENAI_BASE_URL allows pointing to a custom endpoint — if set to an untrusted host it could redirect all prompts to that host.
✓
Persistence & Privilege
The skill is not always-enabled (always: false) and uses the normal autonomous-invocation default. It does not request system-wide configuration or modify other skills. Be aware that an autonomously-invoking skill with access to an LLM API key can run experiments whenever invoked — grant API keys with appropriate limits and monitoring.
Assessment
This skill appears to do what it says: mutate and evaluate skill prompts by calling LLM provider APIs. Before installing or running it: (1) Don't point it at files that contain secrets (API keys, tokens, private data) — the tool will read and send prompt text and test inputs to your configured LLM endpoint. (2) Prefer giving it a provider/API key that has appropriate billing/usage limits or use a test key. (3) If you set OPENAI_BASE_URL, ensure the endpoint is trusted (it controls where your prompts go). (4) Run the script in a sandbox or non-production workspace first and review generated artifacts (results.tsv, SKILL.md.baseline) and mutation history (changelog.md). (5) Review the included autoresearch.py source (already present) if you want to audit exactly what is sent to providers. If you need to optimize prompts that include sensitive content, consider masking or redacting secrets before running autoresearch.

Like a lobster shell, security has layers — review code before you run it.

latestvk979zh8zg7w60fg9vexjpb9mkx83sp4t

License

MIT-0
Free to use, modify, and redistribute. No attribution required.
Termshttps://spdx.org/licenses/MIT-0.html

Runtime requirements

OSmacOS · Linux
Any binpython3, python
Primary envOPENAI_API_KEY

Comments