The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes verification steps and project-appropriate tests during the evaluation and review phases as defined in SKILL.md. These commands are sourced from verification arrays in task slices (tasks/.json) and contract files (contracts/.md) generated during the planning phase.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface during the Research Fanout phase in SKILL.md. External implementation recommendations (ingestion points: research/fanout/*.json) are incorporated into the generator's context as constraints. Boundary markers for this external data are absent, and the skill possesses capabilities to execute commands via the evaluator agent. No sanitization or validation of the research content is specified.

pi-protocol