The Agent Skills Directory

[PROMPT_INJECTION]: The skill employs strong behavioral directives, such as 'This section overrides feature enthusiasm, founder bias, and technical elegance,' to ensure the agent adheres to its 'Product Owner' persona. While these are technically overrides, they are intended behavioral constraints for the skill's logic rather than attempts to bypass system safety filters.
[COMMAND_EXECUTION]: The file PROMOTE.md contains a prompt that instructs an agent to perform file system operations, including reading and writing documentation and skill wrappers within the repository and the ~/.agents/skills/ directory. This is a functional requirement for the skill's distribution and synchronization mechanism.
[DATA_EXFILTRATION]: The skill logic requires reading project-specific files like AGENTS.md, ROADMAP.md, and CONTEXT.md to identify product constraints. This represents a surface for indirect prompt injection (ingesting untrusted data), but it is documented as the primary input for the skill's reasoning process and involves standard, non-sensitive project artifacts.

product-owner