The Agent Skills Directory

[COMMAND_EXECUTION]: Executes shell-based testing and linting tools such as pytest, ruff, and mypy to assess project quality. The skill incorporates several safety guidelines, including mandatory timeouts, budget limits on loops, and non-interactive command flags (< /dev/null), to minimize risks of terminal hangs or system instability. Additionally, it dynamically generates code improvements that are applied and re-verified through the same command-line interface.
[PROMPT_INJECTION]: The skill processes test output and log data to generate code improvements, which presents a surface for indirect prompt injection from codebase content. * Ingestion points: Reads test results and error messages from pytest and other tools across its workflow files (e.g., rules/iteration-workflow.md). * Boundary markers: Explicit delimiters for separating untrusted test logs from the agent's instructions are not defined in the provided workflow. * Capability inventory: Possesses file-writing capabilities (apply_improvements) and shell execution capabilities (run_tests) across multiple files. * Sanitization: No specific mention of filtering or sanitizing test output before it is parsed by the agent for logic generation.

iteration-manager