The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest high-level project objectives and external plan files, which presents a surface for indirect prompt injection.
Ingestion points: The skill reads plan.md (or similar files) produced by other agents and accepts direct user objectives as documented in SKILL.md.
Boundary markers: No explicit boundary markers or "ignore embedded instructions" warnings are defined for the input data in SKILL.md.
Capability inventory: The skill uses the Write tool to create persistent markdown files (task_plan.md, progress.md, findings.md) and generates implementation steps that include shell-based verification commands (e.g., pytest, npm test).
Sanitization: There is no evidence of sanitization, validation, or escaping of the external objective text before it is interpolated into the decomposition logic.

task-decomposition