The Agent Skills Directory

[COMMAND_EXECUTION]: The skill interacts with the local development environment via the doctrine command-line utility, specifically using subcommands like design start, design resume, and slice status to manage design fragments and revisions.\n- [INDIRECT_PROMPT_INJECTION]: The instructions explicitly direct the agent to treat the output of the doctrine design command as its active instruction set ("that output is your instruction: read it and do what it says"). This creates an indirect injection surface.\n
Ingestion points: Output generated by the doctrine CLI tool (SKILL.md).\n
Boundary markers: Absent; the skill lacks specific delimiters to distinguish between tool data and instruction context.\n
Capability inventory: The agent can execute doctrine CLI commands and navigate to other functional routes such as /plan and /knowledge.\n
Sanitization: Absent; there is no evidence of filtering or validation of the tool's output before the agent follows its instructions.

design