maintain-formal-verification-skill
Pass
Audited by Gen Agent Trust Hub on Nov 11, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes external, untrusted data including git commit messages and code diffs to decide how to update verification models. This represents an attack surface where a maliciously crafted commit message could influence the agent's behavior.
- Ingestion points: Reading commit messages and code diffs during the assessment of changes (Step 3 in SKILL.md).
- Boundary markers: None. The instructions do not specify that external content should be wrapped in delimiters or that embedded instructions should be ignored.
- Capability inventory: The skill has the ability to execute shell commands (retrieved from other skills) and write to project files in the .agents/skills directory.
- Sanitization: None. The agent is instructed to read and interpret the messages directly to determine the intent of code changes.
- [COMMAND_EXECUTION]: The skill identifies and executes shell commands defined in other skill files found in .agents/skills or .claude/skills. While this is the intended purpose for running formal verifiers, it involves executing logic defined in external project files which the agent does not directly control.
Audit Metadata