Self Improving 1.1.3
v1.0.0Self-reflection + Self-criticism + learning from corrections. Agent evaluates its own work, catches mistakes, and improves permanently.
⭐ 0· 558·22 current·24 all-time
by@sieyer
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
OpenClaw
Benign
high confidencePurpose & Capability
Name/description (self-reflection, learning from corrections) align with what the skill actually requires: no external services, no credentials, and only local files under ~/self-improving/. There are no unexpected binaries or env vars.
Instruction Scope
SKILL.md instructs the agent to create, read, and update files inside ~/self-improving/ (memory.md, corrections.md, projects/, etc.), and to export ZIPs on demand. This is consistent with the stated purpose. One ambiguity: 'If project detected → preload relevant namespace' is underspecified about detection mechanism (likely session/context-based, but could lead a permissive agent to scan other workspace files). The skill relies on the agent to obey the declared security boundaries (e.g., 'never store credentials'), which are policy statements rather than enforced technical controls.
Install Mechanism
Instruction-only skill with no install spec and no external downloads — lowest-risk install profile.
Credentials
No environment variables, credentials, or external config paths are requested. All declared configPaths point to a single user-scoped directory (~/self-improving/) which matches the skill's purpose.
Persistence & Privilege
The skill is persistent in that it creates and maintains files in the user's home directory and loads memory.md on session start. It does not request always:true, nor does it claim system-wide changes or modify other skills. Users should be aware this creates persistent local state that must be managed or deleted if desired.
Assessment
This skill is internally consistent and appears to do only local memory management. Before installing: 1) Be comfortable with the agent creating ~/self-improving/ and storing preferences/corrections there. 2) Confirm the agent's operating mode (Passive vs Active); choose Passive if you want conservative learning. 3) Don't paste secrets or sensitive third-party data into interactions that might be logged — the skill's policy forbids storing credentials, but that is a behavioral rule, not a technical enforcement. 4) Set filesystem permissions on ~/self-improving/ if the machine is shared, and periodically review/export/delete the files (there are built-in export and forget flows). 5) If you need stronger guarantees (e.g., automatic filtering of sensitive tokens), require technical controls or avoid enabling active learning modes.Like a lobster shell, security has layers — review code before you run it.
latestvk97dcza52wg2m9dnn9k7gqzqyd825zqz
License
MIT-0
Free to use, modify, and redistribute. No attribution required.
Runtime requirements
🧠 Clawdis
OSLinux · macOS · Windows