The Agent Skills Directory

[PROMPT_INJECTION]: The skill employs highly directive and overriding language such as 'ABSOLUTELY MUST', 'not negotiable', and 'not optional' to dictate agent behavior and bypass normal decision-making processes regarding tool usage.- [PROMPT_INJECTION]: It specifically instructs the agent to ignore its own internal logic or 'rationalization' and skip clarifying questions when deciding whether to invoke a skill, which is a common pattern used to bypass safety or logic constraints.- [PROMPT_INJECTION]: The skill establishes an attack surface for Indirect Prompt Injection by mandating the ingestion of external content. 1. Ingestion points: The Skill tool loads content from external SKILL.md files. 2. Boundary markers: No delimiters or 'ignore embedded instructions' warnings are provided for the loaded content. 3. Capability inventory: The agent is instructed to 'follow skill exactly', which may include any tool or command the loaded skill provides. 4. Sanitization: No sanitization or validation of the external skill content is performed before the agent is required to follow its instructions.

using-superpowers