The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data from project specifications and documentation to generate instructions that are subsequently used by other agent skills, creating a potential vector for indirect prompt injection.\n
Ingestion points: The skill reads requirements from a --spec path, spec.md, and project context files including conductor/product.md and conductor/tech-stack.md (SKILL.md).\n
Boundary markers: The instructions do not define delimiters or explicit "ignore instructions" directives to prevent embedded commands in the input files from influencing the plan generation.\n
Capability inventory: The skill possesses file system capabilities to read the input data and write the resulting implementation plans and metadata to the conductor/tracks/ directory (SKILL.md).\n
Sanitization: There are no documented procedures for sanitizing, validating, or filtering the content retrieved from the external specification sources before interpolation into the generated plan.

writing-plans