The Agent Skills Directory

[COMMAND_EXECUTION]: The skill requires the use of shell commands to perform Git operations, including git clone (shallow clone), git commit, and git push to manage decision records in the target repository.
[EXTERNAL_DOWNLOADS]: At session start, the skill clones a remote repository specified by the DECISION_MEMORY_REPO variable to retrieve the active preference set and instructions.
[DATA_EXFILTRATION]: The skill is designed to export session data, including user decisions, rejection reasons, and 'session-local facts' (derived from the local filesystem and tools), to an external repository. While this is the intended functionality, it constitutes a data export to a remote destination.
[PROMPT_INJECTION]: The skill exhibits an Indirect Prompt Injection surface by instructing the agent to adopt instructions from an external source.
Ingestion points: The agent reads and follows behavior rules from an agent-instructions file and preferences.md within the remote DECISION_MEMORY_REPO.
Boundary markers: No explicit boundary markers or 'ignore' instructions are provided for the external content.
Capability inventory: The agent has access to local filesystem/tools ('Facts findable in environment') and Git network operations.
Sanitization: No sanitization is performed on the remote instructions before they are integrated into the agent's operating context.

grilling