Skillv1.0.0

ClawScan security

Agente Conhecimento · ClawHub's context-aware review of the artifact, metadata, and declared behavior.

Scanner verdict

BenignFeb 17, 2026, 11:41 PM

Verdict: benign
Confidence: medium
Model: gpt-5-mini
Summary: The skill's instructions and requested resources are coherent with its stated purpose (logging and promoting learnings to an OpenClaw workspace), but there are a few minor omissions and privacy-related notes you should be aware of before installing.
Guidance: This skill appears to do what it says: append learnings/errors/feature-requests to markdown files in an OpenClaw workspace and optionally promote them to shared workspace files. Before installing or following the manual steps: (1) review any remote repository (the manual install references https://github.com/peterskoett/self-improving-agent.git) before running git clone; (2) confirm you want a hook copied to ~/.openclaw/hooks and enabled (hooks run on session start and create persistent behavior); (3) be aware that using session-sharing commands (sessions_history, sessions_send, etc.) can expose other session transcripts or data—only enable those features if you trust the workspace and its participants; (4) note the SKILL.md references binaries (clawdhub, git, openclaw) but none are declared — ensure those commands are the expected, official tools on your system. If you need higher assurance, ask the author for the hook code or a reviewed install package before enabling.

Review Dimensions

Purpose & Capability: okThe skill claims to capture learnings/errors/feature requests and to write/promote them into an OpenClaw workspace; the SKILL.md exclusively describes creating and appending markdown logs in .learnings and promoting items into workspace files (AGENTS.md, SOUL.md, TOOLS.md, etc.), which matches the stated purpose. Minor inconsistency: the instructions reference external commands (clawdhub, git, openclaw) but the registry metadata lists no required binaries.
Instruction Scope: noteAll runtime instructions are about creating/maintaining local markdown files and optionally enabling an OpenClaw hook. The skill references inter-session operations (sessions_list, sessions_history, sessions_send, sessions_spawn) — these are platform features that can expose session transcripts or send learnings between sessions, so using them has privacy implications. The SKILL.md also suggests cloning a third-party GitHub repo and copying hooks into ~/.openclaw, which modifies the user's workspace/hook directory (expected for this integration but worth reviewing).
Install Mechanism: okThere is no install spec in the registry (instruction-only), so nothing will be automatically downloaded or written by the platform. The documentation suggests optional installation via 'clawdhub install' or a manual 'git clone' from https://github.com/peterskoett/self-improving-agent.git — those are user-run steps and not performed automatically by the skill, but cloning an external repo carries the usual risk and should be reviewed before executing.
Credentials: okThe skill requests no environment variables, no credentials, and no special config paths beyond the OpenClaw workspace (~/.openclaw/workspace and ~/.openclaw/hooks). This is proportionate to a logging/self-improvement skill. No unexplained SECRET/TOKEN/PASSWORD requests are present.
Persistence & Privilege: noteThe skill is not marked 'always' and does not request elevated system privileges. However, it instructs copying/enabling a hook in ~/.openclaw/hooks which would create persistent behavior inside the OpenClaw environment (automatic reminders on session start). Enabling hooks or using session-sharing features grants the skill ongoing presence within your OpenClaw workspace — expected for this kind of integration but something you should explicitly approve.