The Agent Skills Directory

[SAFE]: The skill is designed for standard Python refactoring and includes hard constraints and stop conditions to maintain a narrow scope.
[COMMAND_EXECUTION]: The workflow involves running local development tools such as mypy, pytest, and pre-commit. These are standard tools and the commands are used for code validation within the local repository.
[PROMPT_INJECTION]: The skill does not contain any direct prompt injection, jailbreak attempts, or safety bypass patterns.
[DATA_EXFILTRATION]: No network access or sensitive file exposure patterns were detected. The skill operates on local repository files only.
[PROMPT_INJECTION]: Analysis of the indirect prompt injection surface: * Ingestion points: The skill accepts a module list and worktree name from the task prompt and reads the local pyproject.toml and Python source files. * Boundary markers: Clear 'Scope Rules' and 'Stop Conditions' are present in SKILL.md to restrict the agent's behavior. * Capability inventory: The agent can modify local files and execute shell commands for testing and linting (mypy, pytest, pre-commit). * Sanitization: No explicit sanitization is defined, as the skill operates within a controlled development environment.

typing-exclusion-worker