The Agent Skills Directory

[COMMAND_EXECUTION]: The skill configures PostToolUse and PreToolUse hooks to execute shell scripts (multi_linter.sh, protect_linter_configs.sh) on every file modification within the agent's environment.- [COMMAND_EXECUTION]: The system is designed to spawn claude -p subprocesses to automatically apply code fixes based on linter output, representing a secondary execution layer.- [PROMPT_INJECTION]: The automated fixing workflow creates an indirect prompt injection surface. It feeds linter violations (which include code snippets from project files) into an AI subprocess. If the project files contain malicious instructions disguised as comments or code, they could potentially influence the behavior of the subprocess tasked with fixing the code.- [SAFE]: The skill references well-known development tools such as ruff, uv, biome, and shellcheck for its operations.

plankton-code-quality