The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands such as make test and uv run pytest to identify and verify test fixes. These commands are standard for the intended software development use case and do not involve unusual parameters or privilege escalation.
[INDIRECT_PROMPT_INJECTION]: The skill defines an attack surface where the agent processes potentially untrusted data from test suite outputs.
Ingestion points: The agent reads and analyzes output from make test and pytest (SKILL.md) to identify failure patterns.
Boundary markers: No specific delimiters or instructions are provided to the agent to distinguish between test metadata and potentially malicious content within test failures.
Capability inventory: The skill utilizes file modification capabilities (Edit tool) and shell command execution (make test, pytest, git diff).
Sanitization: There is no explicit sanitization or filtering of the test output before analysis, which is standard for developer-facing tools but creates an inherent surface for indirect injection if tests are run on malicious source code.

test-fixing