The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The script scripts/fetch_thread.py attempts to read ~/.git-credentials to extract GitHub authentication tokens.
[CREDENTIALS_UNSAFE]: Multiple scripts, including scripts/search.py and scripts/relevance_gate.py, read a configuration file at ~/.openclaw/credentials/search.json containing API keys for Grok, Exa, and Tavily.
[INDIRECT_PROMPT_INJECTION]: The skill represents a surface for indirect prompt injection by processing content from arbitrary external sources like Reddit, GitHub, and general search results.
Ingestion points: Untrusted data enters via scripts/fetch_thread.py and scripts/search.py.
Boundary markers: Missing specific delimiters for external content to prevent the agent from executing instructions embedded in search snippets or discussion threads.
Capability inventory: The skill can perform network operations via requests and urllib and follows links recursively through scripts/chain_tracker.py.
Sanitization: Content is cleaned of HTML tags but is not sanitized or filtered for malicious instructional content.
[DYNAMIC_EXECUTION]: scripts/search.py uses the importlib.util module to dynamically load the scripts/fetch_thread.py script at runtime.

search-layer