The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a local script located at ../rex-subagents/scripts/sessions to list and read conversation transcripts. It also uses gh pr create for finalizing changes and git log for checking file history.
[INDIRECT_PROMPT_INJECTION]: The skill has an inherent attack surface for indirect prompt injection because it mines and summarizes instructions from past conversation history.
Ingestion points: Conversation transcripts retrieved via the sessions read command (SKILL.md, Step 1).
Boundary markers: The skill does not define specific delimiters or 'ignore' instructions for the content being mined from history.
Capability inventory: The skill can spawn subagents (subagent spawn), write files (SKILL.md), and interact with GitHub (gh pr create).
Sanitization: No sanitization or validation of the mined text is mentioned before it is used to draft new instructions.

automate-me