The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the execute tool to run ruff for linting and pytest for running test suites. This execution of local code and development tools is the primary function of the skill and is used to provide objective feedback to the agent during code refinement cycles.
[SAFE]: No patterns of prompt injection, data exfiltration, or obfuscation were detected. The skill manages context via local files (/memory/experiment-memory.md) and logs its progress to the /artifacts/ directory. Although the skill interacts with potentially untrusted workspace code during testing, it does so using standard development procedures without any instructions that deviate from its stated purpose.
[COMMAND_EXECUTION]: Regarding the data ingestion surface for indirect prompt injection: 1. Ingestion points: The skill reads existing code and tests from the workspace and context from /memory/experiment-memory.md. 2. Boundary markers: Explicit boundary markers are absent in the prompt templates. 3. Capability inventory: The skill has access to execute (shell commands), write_file, and edit_file. 4. Sanitization: No explicit sanitization or filtering of workspace content is performed before processing. These factors represent a standard attack surface for coding assistants but are not used maliciously in this skill.

experiment-iterative-coder