grace-migrate
Pass
Audited by Gen Agent Trust Hub on Aug 25, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill ingests and processes legacy project documentation (untrusted data) to map requirements and technology into a new format. This creates a surface for indirect prompt injection where instructions hidden within the source files could attempt to influence the agent's behavior.
- Ingestion points: Documentation detection and mapping steps in SKILL.md (Steps 1 and 3).
- Boundary markers: The skill uses structured XML-like tags to define scope and generates a structured migration report (references/migration-report-template.xml) to maintain clear state separation.
- Capability inventory: The agent can execute the grace and git CLI tools and perform file system writes and deletions (Step 9).
- Sanitization: Output is validated using automated linting (grace lint) and status checks (grace status) before any cleanup is proposed.
- [COMMAND_EXECUTION]: The workflow involves executing shell commands for linting, status reporting, and git inspection. It also performs file archiving and deletion. These actions are appropriate for a migration tool and are protected by multiple safety gates, including a mandatory backup outside the cleanup set and explicit human-in-the-loop approvals for both writing new files and cleaning up legacy ones.
Audit Metadata