Back to skill
Skillv7.0.0

ClawScan security

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

Scanner verdict

BenignApr 9, 2026, 12:37 PM
Verdict
Benign
Confidence
medium
Model
gpt-5-mini
Summary
The skill's code, instructions, and requirements are internally consistent with a local, filesystem-only memory search tool; no network calls or unrelated credentials are requested, but it does persist indexed copies of your memory files to disk so review before use.
Guidance
This skill appears to do what it says: local BM25-style search over your memory markdown files, implemented in a pure-Python script with no network behavior or external credentials. Before installing: (1) inspect the full search.py file yourself (and any truncated sections) to confirm there are truly no network calls or unexpected behavior; (2) review the files under ~/.openclaw/workspace (MEMORY.md and memory/*.md) for sensitive information because the skill will read them and store an index (~/.openclaw/memory_bm25_index.json); (3) if you want to limit exposure, run the script on a copy or restrict MEMORY_PATHS to a curated directory; (4) remove the index/cache files if you uninstall the skill or after testing. Confidence is medium because a portion of the source shown was truncated — verify the remainder before trusting it fully.

Review Dimensions

Purpose & Capability
okName/description (local BM25-based memory search) match the files and runtime behavior: the script reads markdown memory files under ~/.openclaw/workspace, builds a local JSON index, and performs BM25 + multi-signal reranking. No unrelated credentials, binaries, or services are requested.
Instruction Scope
noteSKILL.md directs the agent to run the included Python script and to read/write index files under ~/.openclaw. That is appropriate for a local memory search skill. Note: the skill will read any files matching the configured MEMORY_PATHS (default ~/.openclaw/workspace/MEMORY.md and ~/.openclaw/workspace/memory/*.md), so it has access to all data in those paths — which is expected but could expose sensitive local content via search results.
Install Mechanism
okNo install spec; this is instruction + inline script only. No downloads or external packages are pulled — the script claims to use only Python stdlib and included code appears to implement that.
Credentials
okThe skill requests no environment variables, no external API keys, and no unrelated configuration paths. The declared scope (workspace memory files and local index/cache) is proportional to its stated functionality.
Persistence & Privilege
notealways:false and normal invocation. The script creates/reads persistent files in the user home: ~/.openclaw/memory_bm25_index.json and a token/cache path (~/.openclaw/memory_bm25_token_cache.json). These files store derived copies/indices of your memory files and will persist on disk until removed — consider this when handling sensitive content.