writing-plans
Pass
Audited by Gen Agent Trust Hub on Jun 20, 2026
Risk Level: SAFE
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill directs the agent to perform reconnaissance by reading project configuration files (e.g.,
package.json,justfile,noxfile,CIconfigs) to identify valid build and test commands. This creates a surface for indirect prompt injection if the local repository contains malicious instructions within these configuration files. - Ingestion points: Project configuration files and agent-specific documentation like
CLAUDE.mdorAGENTS.mdlocated in the user's repository. - Boundary markers: The skill does not define specific delimiters for the content ingested from these files.
- Capability inventory: The skill instructions imply that the agent should verify discovered commands (via shell execution) to ensure they work as expected before including them in a plan.
- Sanitization: No specific sanitization logic is provided for the commands extracted from external configuration files.
- [COMMAND_EXECUTION]: During the 'Recon' phase, the skill instructs the agent to verify exact build, test, and lint commands from the repository's configuration rather than guessing them. This involves executing these shell commands to confirm their validity, which is a necessary step for its intended purpose of creating accurate implementation plans.
Audit Metadata