test-audit
Pass
Audited by Gen Agent Trust Hub on Oct 4, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill has an attack surface for indirect prompt injection as it ingests untrusted data from the local repository.
- Ingestion points: Reads test files, production source code, commit history, and AGENTS.md files (SKILL.md and campaign.md).
- Boundary markers: The skill explicitly mandates a read-only discovery phase and requires user approval before performing any modifications.
- Capability inventory: The skill can modify local files (source and tests) and execute shell commands for test validation and git operations.
- Sanitization: There is no automated sanitization of the ingested code content; the skill relies on the agent's internal safety filters and mandatory human review of proposed changes.
- [COMMAND_EXECUTION]: The skill utilizes local development tools such as test runners, git, and formatters. These commands are standard for the described purpose and are used within a validation context after user approval.
Audit Metadata