The Agent Skills Directory

[PROMPT_INJECTION]: The skill uses high-pressure directives like '', 'ABSOLUTELY MUST', and 'This is not negotiable' to override the agent's discretion regarding tool use. This behavioral override is used to enforce a specific workflow methodology (ACP).
[PROMPT_INJECTION]: The 'Red Flags' section instructs the agent to ignore internal reasoning patterns, such as the desire to ask clarifying questions or perform initial codebase exploration, and instead prioritize immediate skill invocation. This is a targeted modification of the agent's standard procedural logic.
[SAFE]: The skill incorporates a robust safety mechanism under 'Instruction Priority', which explicitly mandates that the user's direct requests and project-specific configuration files (e.g., CLAUDE.md) always take precedence over the skill's instructions. This ensures that the agent remains responsive to user control and safety constraints.

acp-loader