The Agent Skills Directory

[PROMPT_INJECTION]: The skill facilitates an indirect prompt injection surface by ingesting data from phase-2-assignments.md and interpolating it into prompts for the Agent tool. * Ingestion points: phase-2-assignments.md (referenced in Step 1 and Step 3). * Boundary markers: The skill instructions require prompts to be 'fully self-contained' and include a 'handoff block', but it lacks explicit technical delimiters (such as XML tags or unique sequence markers) to wrap the untrusted content. * Capability inventory: The skill has access to Read, Write, Edit, Bash, Glob, Grep, and Agent tools. * Sanitization: The process implements a mandatory 'no-pingpong gate' using a pingpong-detector agent to identify and block problematic or recursive assignments before they are dispatched.

coliseum-dispatch