The Agent Skills Directory

[PROMPT_INJECTION]: The skill ingests user work descriptions to generate task plans. This creates a standard surface for indirect prompt injection where adversarial user input could theoretically influence the generated instructions for subagents. However, this is inherent to the skill's orchestration purpose and no explicit malicious override patterns were detected.
[COMMAND_EXECUTION]: The skill generates executable shell commands within the 'acceptance' field of the task JSON. These commands are used to verify task completion (e.g., running tests). This is a core feature of the skill's planning logic and is used for quality assurance rather than unauthorized system access.

intent-router