The Agent Skills Directory

[PROMPT_INJECTION]: The skill employs aggressive, high-pressure language (e.g., "YOU ABSOLUTELY MUST," "YOU DO NOT HAVE A CHOICE," "This is not negotiable") to force the agent to prioritize skill invocation. It explicitly instructs the agent that "Superpowers skills override default system behavior," which is a direct attempt to modify the agent's core decision-making process.
[PROMPT_INJECTION]: The skill mandates that the agent check for and invoke skills "BEFORE any response or action," including clarifying questions. This pattern forces tool execution into the loop before the agent can interact with the user, potentially enabling automated workflows that bypass manual confirmation.
[COMMAND_EXECUTION]: The reference files (references/codex-tools.md, references/gemini-tools.md) provide mappings for shell execution and file system manipulation tools. While these are standard capabilities, the meta-skill's insistence on mandatory invocation increases the surface area for automated command execution.
[INDIRECT_PROMPT_INJECTION]: The skill creates an attack surface by requiring the agent to load and follow external skills if there is even a "1% chance" they apply.
Ingestion points: The Skill tool (or activate_skill) is used to load external content based on broad triggers defined in this skill.
Boundary markers: The skill includes a <SUBAGENT-STOP> block and explicit instruction priority rules (User > Skill > System).
Capability inventory: The mapped tools include run_shell_command, replace (file edit), write_file, and web_fetch across the reference files.
Sanitization: There is no mention of sanitizing or validating the content of the external skills being invoked.

using-superpowers