The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes a comprehensive suite of gbrain schema CLI tools and standard git commands. These are used for administrative tasks such as linting schemas, querying database statistics, and managing version control for schema packs within the ~/.gbrain/ directory.
[DATA_EXFILTRATION]: In the commitment phase, the skill executes gbrain git commit-push to synchronize local schema changes with a remote repository. This network operation is scoped specifically to the pack.json file within the local schema-pack repository, representing a legitimate data synchronization workflow rather than unauthorized exfiltration.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection as it scans untrusted user content (the document corpus) to propose new schema types.
Ingestion points: Document corpus scanning during the detect and suggest phases in SKILL.md.
Boundary markers: The instructions do not define explicit delimiters or instructions for the agent to ignore potentially malicious content within the analyzed corpus.
Capability inventory: The skill can write configuration files to ~/.gbrain/schema-packs/, perform bulk updates to database records via sync --apply, and execute network transfers via git push.
Sanitization: There is no explicit mention of sanitizing or filtering instructions from the document corpus before they are processed by the suggestion heuristic or the LLM.

schema-author