The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses system inspection commands including ps, top, and cat /proc/* to analyze the state of the environment and identify causes of process failure.
[DATA_EXFILTRATION]: The skill can export diagnostic findings to GitHub via the mcp__github__issue_write tool. This functionality is intended for bug reporting and includes a requirement for user confirmation.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it processes runtime logs that may contain untrusted data. Ingestion points: Log files such as .gsd/activity/*.jsonl and .gsd/journal/*.jsonl are read and analyzed. Boundary markers: The instructions lack specific delimiters or directions to ignore embedded commands within the log files. Capability inventory: The skill has permissions to write to the local filesystem and call external GitHub API tools. Sanitization: No sanitization or filtering of the ingested log content is performed before analysis.

forensics