The Agent Skills Directory

[SAFE]: The file acts as a documentation index for the agent, describing its own capabilities and tools without including executable payloads or malicious overrides.
[COMMAND_EXECUTION]: The skill documents the availability of powerful system tools including bash, exec, and process. These are presented as standard built-in capabilities for administrative and development tasks.
[DATA_EXFILTRATION]: Documents tools capable of reading files (read) and performing network requests (fetch, web_search). While these provide the technical surface for data movement, no malicious exfiltration patterns were detected in the instructions.
[PROMPT_INJECTION]: The skill documents the agent-browser and social-cli tools which ingest untrusted data from external websites and social media platforms.
Ingestion points: Browser automation via CDP (agent-browser), web fetching (fetch), and social platform scrapers (social-cli).
Boundary markers: None explicitly defined in this index file.
Capability inventory: Includes shell access (bash), file operations (read/write), and the ability to load additional code (skill).
Sanitization: None described within the index documentation.

skill-index