The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes user-provided feature descriptions to generate plan documents, which introduces a potential surface for indirect prompt injection. \n- Ingestion points: High-level feature ideas provided by the user are used to construct WORKFLOW_*.md files. \n- Boundary markers: The skill mandates procedural checks against AGENTS.md but does not specify technical delimiters for user input in its interpolation logic. \n- Capability inventory: Actions include creating markdown files, monitoring repository status with git, and invoking verification skills like compiler. \n- Sanitization: The instructions do not define specific sanitization routines for the user's initial input. \n- [COMMAND_EXECUTION]: The skill incorporates standard developer commands to maintain environment synchronization and verify code changes. \n- Evidence: The skill requires executing git status .agents/skills/ to synchronize documentation for expert skills. \n- Evidence: The generated checklists mandate the use of gradle_sync and the compiler verification suite as part of the development lifecycle.

workflow-feature