The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses timeout_run 1 bd current to resolve an internal context identifier (bead ID). This is a standard local operation for managing agent state and does not involve external network calls or dangerous elevated privileges.
[DATA_EXPOSURE_AND_EXFILTRATION]: The skill writes information provided by the user to local files in the .agents/learnings/ directory. No sensitive files are accessed, and no data is transmitted to external servers.
[INDIRECT_PROMPT_INJECTION]: The skill ingests user input and writes it to markdown files. While it lacks explicit sanitization, the risk is minimal as the output is stored in a dedicated directory for documentation rather than being executed as code.

retro