ClawHub

Self-Improving + Proactive Agent

v1.2.16

Self-reflection + Self-criticism + Self-learning + Self-organizing memory. Agent evaluates its own work, catches mistakes, and improves permanently. Use when...

⭐ 927· 155k·1.7k current·1.8k all-time
byIván@ivangdavila
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
The skill claims to build a local, self-improving memory and the SKILL.md only requires creating/reading files under ~/self-improving/ and updating workspace AGENTS.md/SOUL.md/HEARTBEAT.md non-destructively. No unrelated binaries, cloud credentials, or network tokens are requested. The declared configPaths match the described purpose.
ℹ
Instruction Scope
Runtime instructions tell the agent to read/write ~/self-improving/*, load memory.md on session start, scan for changes, and append structured logs (corrections, reflections, heartbeats). They also instruct non-destructive edits to workspace files (AGENTS.md, SOUL.md, HEARTBEAT.md) and to run simple shell commands (mkdir, find) as part of setup. This behavior is consistent with the purpose but you should be aware the skill will read and write files in your home directory and may modify workspace files unless you decline that step.
✓
Install Mechanism
There is no install spec and no code files — the skill is instruction-only, which is low risk. The only network-related operation mentioned is an optional, consent-gated 'clawhub install proactivity' step; the SKILL.md explicitly says to only run that after the user agrees.
✓
Credentials
The skill requires no environment variables or credentials. It will store user-provided corrections and preferences in local files; boundaries.md explicitly forbids storing secrets/credentials and describes consent and audit flows, which is appropriate for the stated functionality.
ℹ
Persistence & Privilege
The skill persists state locally under ~/self-improving/ (memory.md, corrections.md, heartbeat-state.md, etc.). always:false and normal model invocation are used. Persisting learned preferences and automatic heartbeat maintenance is expected for this class of skill, but it does create lasting local state and can modify workspace steering files non-destructively — worth awareness before enabling.
Assessment
This skill is an instruction-only local memory/learning system and appears to do what it says: create and maintain ~/self-improving/ and update small workspace steering files non-destructively. Before installing or enabling it: (1) Confirm you are comfortable letting the agent read and write files under ~/self-improving/ and make limited edits to AGENTS.md/SOUL.md/HEARTBEAT.md; (2) Back up AGENTS.md and SOUL.md if you want an easy rollback; (3) Note the skill will not request credentials, but it will store user-provided corrections and preferences—review boundaries.md and ensure it meets your privacy expectations; (4) Only allow the optional 'Proactivity' companion to be installed if you review that skill first (it runs a network install); (5) Verify the skill's kill-switch and export/wipe flows work as promised (ask it to export and then to wipe on a test run) and limit file permissions on ~/self-improving/ if you want stricter local access control.

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

latestvk972fcwc6mnswg94vwpp43ynv182rntp

License

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

Runtime requirements

🧠 Clawdis
OSLinux · macOS · Windows

Comments