The Agent Skills Directory

[SAFE]: No malicious patterns or security risks were detected. The skill utilizes standard software development workflows and tools (make, git, pytest) to fulfill its stated purpose of fixing tests.\n- [INDIRECT_PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection because it relies on processing external data (test results) to determine its actions.\n
Ingestion points: Test output and error logs generated by make test and pytest.\n
Boundary markers: Absent; there are no specific delimiters or instructions for the agent to ignore potentially malicious directives embedded within failing test output.\n
Capability inventory: The agent is instructed to use the Edit tool to modify files and execute commands such as make, pytest, and git.\n
Sanitization: Absent; the skill directs the agent to read and analyze error patterns directly from test output without filtering.\n
Note: This behavior is a fundamental requirement for the skill's primary function and does not represent a malicious intent.

test-fixing