improve-test-cases
Pass
Audited by Gen Agent Trust Hub on Oct 23, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests and processes untrusted data from the local environment, creating a potential surface for instructions embedded in test files to influence agent behavior.
- Ingestion points:
SKILL.md(Step 1) instructs the agent to locate and read markdown files (.test.mdor.md) from various project directories. - Boundary markers: The instructions do not define specific delimiters or "ignore" markers to prevent the agent from obeying commands that might be hidden inside the test case content.
- Capability inventory:
SKILL.md(Step 5) provides the agent with the capability to overwrite original files in the local directory structure. - Sanitization: There is no mention of sanitization or filtering of the file content before it is analyzed and used to generate updates.
- Mitigation: The skill includes a human-in-the-loop control in
SKILL.md(Step 3), which requires the agent to show a summary of issues and wait for explicit user confirmation before modifying any files.
Audit Metadata