The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to read and execute instructions from a plan file, which introduces a potential surface for indirect prompt injection if the file contains malicious data. * Ingestion points: The agent is instructed to read a plan file in Step 1 of the workflow (SKILL.md). * Boundary markers: There are no explicit delimiters or instructions provided to ignore or isolate commands embedded within the plan data. * Capability inventory: The skill orchestrates task execution, verification runs, and git operations through integrated workflow skills (superpowers:using-git-worktrees, superpowers:finishing-a-development-branch), which involve file system and shell access. * Sanitization: No automated sanitization or schema validation of the plan content is described; however, the skill mandates a critical review by the agent and periodic reporting for human feedback (Step 3) to reduce the risk of executing unintended instructions.

executing-plans