The Agent Skills Directory

[COMMAND_EXECUTION]: The skill execution includes running local Python scripts: profile_store.py for configuration management and validate_skill.py for ensuring skill standards are met.
[PERSISTENCE]: The profile_store.py script manages a local JSON configuration file (stored in directories like ~/.config/agent-skills/) to persist user and skill-specific data between sessions.
[DATA_EXFILTRATION]: While the skill manages local configuration data, it explicitly implements a validation mechanism that blocks the recording of sensitive fields such as 'token', 'secret', 'password', and 'api_key' to the filesystem.
[INDIRECT_PROMPT_INJECTION]: The skill is intended to process external project data and user input, creating a surface for potential instruction injection.
Ingestion points: User-provided background information and project materials read during the workflow in SKILL.md.
Boundary markers: Instructions require the agent to distinguish between user-supplied facts and internal assumptions.
Capability inventory: The skill has the ability to write specific, non-sensitive configuration data to the local profile via the profile_store.py script.
Sanitization: A sensitive key blacklist is applied to all data segments in profile_store.py before it is written to the profile store.

lov-brainstorm