prd-to-tasks
Pass
Audited by Gen Agent Trust Hub on Mar 27, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands such as
npm test,lint, andbuildto verify phase completion. These commands are pulled from the local project's environment. - Evidence: Found in
SKILL.md(Phase 3.5) andreferences/task-schema.md(Phase Completion Task), which direct the agent to run project-specific check scripts. - [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by reading and processing user-provided content from
prd.mdto drive its logic. - Ingestion points:
plans/<name>/prd.mdandplans/<name>/tasks.md. - Boundary markers: The instructions lack explicit delimiters to separate untrusted data from core instructions or to warn the agent to ignore embedded commands.
- Capability inventory: The agent has permissions to read and write local files and execute shell commands discovered in the project.
- Sanitization: No sanitization or validation of the input content is performed before it is analyzed to generate task lists.
Audit Metadata