← Back to Skills Marketplace
Self-Improving Science
by
José I. O.
· GitHub ↗
· v1.1.0
· MIT-0
105
Downloads
0
Stars
0
Active Installs
2
Versions
Install in OpenClaw
/install self-improving-science
Description
Captures learnings, experiment issues, and methodology corrections for continuous improvement in scientific research and ML workflows. Use when: (1) Data lea...
Usage Guidance
This skill appears to do what it says: local reminders and simple, opt-in scripts to log experiment learnings. Before enabling: 1) Review the hook scripts (activator.sh, error-detector.sh) so you’re comfortable they only print reminders and scan local tool output (they do not exfiltrate data). 2) Be mindful when using cross-session features (sessions_history, sessions_send) — those can surface other session transcripts or findings; avoid sending raw sensitive data. 3) Don’t write proprietary data, PII, API keys, or raw datasets into the .learnings/ files; follow the SKILL.md guidance to redact. 4) If you install hooks, prefer the minimal activator-only setup or add matcher filters to reduce noise. 5) If you plan to use the manual git install, inspect the referenced GitHub repo before cloning. If you want a tighter review, provide the remote repository URL or confirm whether the repo owner is trusted.
Capability Analysis
Type: OpenClaw Skill
Name: self-improving-science
Version: 1.1.0
The 'self-improving-science' skill bundle is a legitimate tool designed to help researchers and data scientists log experiment issues, methodology corrections, and feature requests. It includes bash scripts (activator.sh, error-detector.sh, extract-skill.sh) and OpenClaw hooks (handler.js/ts) that facilitate the capture of research metadata and the promotion of learnings into structured artifacts like model cards or checklists. The scripts perform standard file operations with appropriate path validation to prevent directory traversal, and the instructions explicitly advise the agent against logging sensitive data such as API keys or PII.
Capability Tags
Capability Assessment
Purpose & Capability
Name/description (capture learnings, promote to docs/checklists) align with the provided scripts and hook handlers. Included scripts (activator, error-detector, extract-skill) and hook handlers implement reminders, local error-detection, and skill scaffolding which are coherent for this purpose.
Instruction Scope
Runtime instructions and scripts are scoped to creating/maintaining .learnings/ logs, injecting a virtual reminder on agent bootstrap, and scanning tool output for error patterns. The docs also instruct using OpenClaw inter-session APIs (sessions_history, sessions_send, sessions_spawn) to share learnings across sessions — this is consistent with the skill goal but increases risk of exposing cross-session transcripts or learnings if used carelessly. The error-detector reads the CLAUDE_TOOL_OUTPUT environment variable to look for patterns; it does not transmit that content elsewhere, but it does inspect possibly-sensitive tool output.
Install Mechanism
No install spec is provided (instruction-only install), which is the lowest risk category. Manual install examples reference a GitHub repo URL; no arbitrary downloads, shorteners, or archive extraction are used by the skill itself.
Credentials
The skill declares no required env vars or credentials. One script (error-detector.sh) reads CLAUDE_TOOL_OUTPUT (a runtime variable supplied by the host agent) but that env var is not listed in requires.env — this is expected for hook scripts but worth noting because the script inspects tool output which can contain sensitive information if the agent/tool emits it.
Persistence & Privilege
The skill is opt-in: hooks and scripts must be enabled/installed by the user. always:false and no indication of modifying other skills or system-wide configs. Hook handler injects a virtual reminder file at bootstrap; it does not persist credentials or enable itself without user action.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install self-improving-science - After installation, invoke the skill by name or use
/self-improving-science - Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.1.0
**Version 1.1.0**
- Added stackability contract for multi-skill installations.
- Added namespaced logging guidance (`.learnings/science/`) for coexistence with other skills.
- Added required `Skill: science` metadata field and cross-skill precedence/ownership rules.
- Clarified hook arbitration model (single dispatcher, dedupe, rate limiting).
v1.0.0
- Initial release of the self-improving-science skill for logging experiment learnings, issues, and methodology corrections in scientific and ML workflows.
- Provides structured Markdown log files (`LEARNINGS.md`, `EXPERIMENT_ISSUES.md`, `FEATURE_REQUESTS.md`) under a `.learnings/` directory.
- Includes setup instructions and quick-reference actions for different experiment situations.
- Supports promotion of key findings to checklists, data governance documents, and model cards.
- Offers integration guidance for OpenClaw and generic agent environments.
Metadata
Frequently Asked Questions
What is Self-Improving Science?
Captures learnings, experiment issues, and methodology corrections for continuous improvement in scientific research and ML workflows. Use when: (1) Data lea... It is an AI Agent Skill for Claude Code / OpenClaw, with 105 downloads so far.
How do I install Self-Improving Science?
Run "/install self-improving-science" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is Self-Improving Science free?
Yes, Self-Improving Science is completely free, licensed under MIT-0. You can download, install and use it at no cost.
Which platforms does Self-Improving Science support?
Self-Improving Science is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created Self-Improving Science?
It is built and maintained by José I. O. (@jose-compu); the current version is v1.1.0.
More Skills