test-fixing
Pass
Audited by Gen Agent Trust Hub on Oct 29, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill relies on analyzing the output from test suites to perform code modifications, which creates a potential surface for indirect prompt injection from the code under test.
- Ingestion points: Output and error streams from
make testandpytestcommands as described in the 'Initial Test Run' and 'Systematic Fixing Process' sections of SKILL.md. - Boundary markers: None; the agent processes raw output from the shell commands.
- Capability inventory: The skill uses file-writing capabilities (Edit tool) and shell command execution (
make,uv,pytest). - Sanitization: No procedures are described for sanitizing or filtering external command output before it is used to inform code changes.
- [COMMAND_EXECUTION]: The skill requires the agent to execute shell commands such as
make testanduv run pytest. These are standard tools for the task but represent a capability to run subprocesses based on skill instructions.
Audit Metadata