The Agent Skills Directory

[COMMAND_EXECUTION]: The skill relies on executing several local command-line tools and scripts including ./scripts/skill-runs for fetching run history, gh for interacting with GitHub Pull Requests, and ./notify for sending status updates.
[DYNAMIC_EXECUTION]: The skill invokes a local Python script (scripts/skill_health_recovery.py) to process specific recovery logic, passing variable state data as arguments to the script.
[INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection because it reads and processes data from local log files and issue reports. Ingestion points: Reads files from memory/logs/ and memory/issues/. Boundary markers: Uses specific regex signatures and structured YAML frontmatter in issue files to delimit data. Capability inventory: Performs file writes to memory/issues/ and output/ directories, and executes subprocesses via ./notify, gh, and python3. Sanitization: Employs classification rules based on metrics and specific status flags rather than direct interpretation of free-text input.

skill-health