The Agent Skills Directory

[COMMAND_EXECUTION]: The skill requires the python3 binary and uses the pandas library to execute safety management logic and calculate metrics as defined in SKILL.md.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted project data that could contain malicious instructions embedded within safety records.
Ingestion points: According to instructions.md, the agent accepts data in CSV, Excel, and JSON formats, as well as direct user input for hazard and incident descriptions in SKILL.md.
Boundary markers: The skill does not define specific delimiters to separate untrusted data from system prompts.
Capability inventory: The skill is granted filesystem permissions via claw.json to manage project-specific safety data storage.
Sanitization: There is no evidence of explicit input sanitization or validation for the text-based data processed by the agent.

safety-inspection