The Agent Skills Directory

[COMMAND_EXECUTION] (SAFE): The skill utilizes standard shell commands (ls, cat, grep, head) to manage and read its local documentation cache and search index. These operations are restricted to the ~/clawd/data/ directory and are used for legitimate caching purposes.- [EXTERNAL_DOWNLOADS] (SAFE): Content is retrieved from https://docs.clawd.bot/ via the web_fetch tool. The downloads are limited to documentation in markdown format from a domain specific to the tool's purpose.- [DATA_EXFILTRATION] (SAFE): Analysis of the fetch and read patterns shows no attempts to access or transmit sensitive system files or credentials. Network activity is limited to the documentation server.- [PROMPT_INJECTION] (SAFE): No instructions designed to bypass safety filters, extract system prompts, or override agent behavior were found.- [Indirect Prompt Injection] (SAFE): Evaluation of the attack surface:
Ingestion points: Data is ingested through web_fetch from the documentation URL and cat from the local snippet cache.
Boundary markers: None present; the agent is not explicitly instructed to disregard instructions found within the documentation content.
Capability inventory: Includes shell commands (ls, cat, grep, head) and the web_fetch tool.
Sanitization: Absent; paths are constructed using string interpolation.
Risk Assessment: The risk is minimal as the data source is a specific, presumably trusted documentation domain.

clawd-docs-v2