The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from markdown plan files and uses it to drive subagent behavior. The Task Prompt Template in Step 3 interpolates raw task content (Description, Acceptance Criteria, Validation) into the subagent's prompt without using protective delimiters or 'ignore embedded instructions' warnings for that specific content.\n- Ingestion points: The skill reads user-specified markdown files (e.g., plan.md) in Step 1 and Step 2 to extract task details.\n- Boundary markers: Delimiters are absent; content is directly inserted into the Task Prompt Template.\n- Capability inventory: Subagents are granted capabilities to explore the environment, read files, and implement changes to all acceptance criteria (Step 3, Instructions 1, 2, and 4).\n- Sanitization: There is no evidence of sanitization, validation, or filtering of the content extracted from the plan files.

parallel-task