The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from PRD and Tech Spec files to generate implementation tasks, creating a risk that malicious instructions within those documents could be adopted by the agent.
Ingestion points: SKILL.md Step 2 reads ./prds/prd-[feature-slug]/prd.md and techspec.md into the agent's context.
Boundary markers: Absent. The instructions do not specify delimiters or provide warnings to ignore embedded instructions within the source files.
Capability inventory: The skill can write multiple files to the local file system (tasks.md, [num]_task.md) and execute file deletion commands.
Sanitization: Absent. There is no logic to filter or validate content extracted from the PRD/Tech Spec before it is used to generate the task breakdown.
[COMMAND_EXECUTION]: The skill explicitly instructs the agent to use shell commands for file system management.
Evidence: SKILL.md Step 4.10 directs the agent to "delete it with rm via Bash" for any incorrectly named files (e.g., 1.0_task.md) detected in the tasks directory. While scoped to the specific directory, this grants the agent permission to perform file deletions via the shell.

do-create-tasks