repair-proof
Pass
Audited by Gen Agent Trust Hub on Sep 22, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external data including 'source contracts', 'proof results', and 'matching candidates'. This creates a potential surface for indirect prompt injection if these external files contain malicious instructions.
- Ingestion points: The workflow involves reading external source contracts and proof result files as described in
SKILL.md. - Boundary markers: The skill includes guardrails requiring the agent to confirm exact contract revisions and captured text match the current state, which provides a layer of validation against unauthorized changes.
- Capability inventory: The agent is authorized to perform scoped file editing and execution of 'focused checks' (shell commands).
- Sanitization: While the skill mandates revision verification, it does not explicitly describe sanitization of the contract content itself.
- [COMMAND_EXECUTION]: The workflow requires the agent to 'Run the focused check' as part of the repair validation step. This implies the execution of local shell commands to verify the code changes.
Audit Metadata