The Agent Skills Directory

[PROMPT_INJECTION]: The skill employs aggressive, imperative framing (e.g., "ABSOLUTELY MUST", "not negotiable", "not optional") and directives to "not rationalize" designed to override the agent's default logic and internal constraints.
[PROMPT_INJECTION]: Instructions specifically target and disable the agent's standard interaction patterns, such as asking for clarification or gathering context, by mandating tool execution as the highest priority.
[PROMPT_INJECTION]: The skill instructs the agent to prioritize the skill's workflow instructions over user instructions, potentially leading to unauthorized or unexpected actions based on external content.
[PROMPT_INJECTION]: The skill creates a surface for indirect prompt injection by mandating the automatic invocation and strict adherence to content loaded via the Skill tool.
Ingestion points: External skill files loaded via the Skill tool (SKILL.md).
Boundary markers: Absent; instructions explicitly state to "follow skill exactly" without validation or isolation.
Capability inventory: Mandates use of Skill and TodoWrite tools (SKILL.md).
Sanitization: Absent; no validation or filtering of the ingested skill content is provided.

using-superpowers