planning
Pass
Audited by Gen Agent Trust Hub on Oct 27, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: No security vulnerabilities were detected in the skill's instructions or scripts. The workflow focuses on task planning and state persistence using local files. The skill also incorporates security best practices by recommending threat model reviews for tasks involving authentication or external APIs.\n- [INDIRECT_PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface due to its requirement to read and process repository content during the planning phase.\n
- Ingestion points: Repository source code, test files, and local plan artifacts such as .plan/task_plan.md and docs/plans/.\n
- Boundary markers: Uses structured markdown templates and clear headers to delimit sections within the generated planning documents.\n
- Capability inventory: Executes a bundled local shell script (init-plan.sh) for environment scaffolding, performs file writes to the local .plan/ directory and .gitignore, and supports task dispatch to subagents.\n
- Sanitization: The instructions do not define specific sanitization or escaping procedures for external content before its inclusion in the generated plan artifacts.
Audit Metadata