new-self-improving
v1.0.1Log corrections, errors, feature requests, and recurring patterns into structured workspace learning files, then promote stable patterns into tiered memory a...
⭐ 0· 48·0 current·0 all-time
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
OpenClaw
Benign
high confidencePurpose & Capability
The name/description (self-improving memory capture and promotion) aligns with the included files and the single init script. No extraneous environment variables, binaries, or network installs are requested. Creating and managing `.learnings/` and `.self-improving/` is proportionate to the stated goal.
Instruction Scope
SKILL.md instructs the agent to read, append, search, and promote entries across workspace files (raw learnings, review queue, HOT, domains, and, when promoted, core workspace files such as SOUL.md/AGENTS.md/TOOLS.md/MEMORY.md). That behavior is consistent with the purpose but inherently enables permanent edits to the agent's workspace. The instructions explicitly forbid storing secrets and limit scope to local/private workspaces; however, those constraints are procedural (not enforced), so users should be aware the skill will modify workspace files if promotions occur.
Install Mechanism
No install spec and only one small shell init script are included. The script copies shipped templates into workspace directories and is safe to run multiple times; it does not download or execute remote code, and it does not write to unexpected system locations.
Credentials
The skill requests no environment variables or credentials. The SKILL.md explicitly instructs not to store secrets, tokens, or sensitive health/third-party private data. Required access (filesystem in the workspace) is appropriate for the feature.
Persistence & Privilege
The skill persistently creates and appends to files inside the workspace and can promote entries into permanent workspace files (SOUL.md, AGENTS.md, etc.). It does not request always:true or modify other skills' configs. Still, because promotions change core workspace files (which affect agent behavior), users should be aware of and control when promotions happen (manual review or restricted autonomy).
Assessment
This skill is internally coherent and appears to do what it says: create local learning directories, log corrections/errors/feature-requests, review recurring items, and promote stable patterns to higher-tier workspace files. Before installing: (1) Back up any existing SOUL.md / AGENTS.md / TOOLS.md / MEMORY.md or other workspace files the skill might append to. (2) Run the init script in the intended workspace (bash scripts/init-workspace.sh <path>) — it is idempotent and only copies local templates. (3) Decide whether you want the agent to promote entries automatically or only after manual review; if you prefer manual control, avoid enabling autonomous promotions in the agent. (4) Verify the agent redaction policy and avoid logging sensitive data into learnings; although the SKILL.md forbids storing secrets, that is a policy, not a technical enforcement. (5) If you are concerned about automatic modifications to core behavior files, restrict the skill's use to a private sandbox workspace first to observe its actions.Like a lobster shell, security has layers — review code before you run it.
latestvk97bs8a06ps7n3n4w7ydgvzw35849sce
License
MIT-0
Free to use, modify, and redistribute. No attribution required.
Runtime requirements
🧠 Clawdis