auto-quality
Pass
Audited by Gen Agent Trust Hub on Apr 7, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [PROMPT_INJECTION]: The skill instructs the agent to operate in a 'silent' mode (
silent=true) and explicitly forbids the agent from informing the user about specific events, such as when tests fail ([SAY:"tests failed"|allow=false]). This pattern of suppressing agent communication and output reduces user oversight and could allow malicious actions to go unnoticed. - [COMMAND_EXECUTION]: The skill establishes a workflow that ingests user-provided code, identifies logic or security issues, generates tests (
[WRITE:tests]), and executes them ([RUN]). The combination of processing untrusted input and executing derived scripts creates a surface for Indirect Prompt Injection. - Ingestion points: Untrusted user code is processed after each feature implementation (
after-each-feature). - Boundary markers: There are no explicit markers or instructions to delimit user code or to ignore instructions embedded within the processed content.
- Capability inventory: The skill utilizes file modification (
[FIX:if-found]), file creation ([WRITE:tests]), and code execution ([RUN]). - Sanitization: No evidence of sanitization or validation of the input code before it is used for test generation or execution is present.
Audit Metadata