The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the terminal tool to execute shell commands for gathering context and verifying changes.
Evidence: The process involves running git diff to identify changes, git blame to understand code history, and various linter or type-check commands found in the repository.
[DYNAMIC_EXECUTION]: The skill instructs the agent to execute project-specific tests to ensure that refactoring efforts do not introduce regressions.
Evidence: Phase 3, Step 5 explicitly directs the agent to "run the project's targeted tests for the touched files" and "re-run any linter/type check the repo uses."
[INDIRECT_PROMPT_INJECTION]: The skill has a surface for indirect prompt injection as it ingests and processes untrusted code diffs and file content from the local environment.
Ingestion points: The skill captures git diff output and reads arbitrary files via read_file and search_files (SKILL.md, Phase 1 & 2).
Boundary markers: The skill does not explicitly define boundary markers for the diff content but instructs the agent to discard false positives and categorize findings by risk.
Capability inventory: The skill has access to terminal execution, file writing, and task delegation tools.
Sanitization: Verification is performed via automated tests and human review is required for any findings categorized as 'RISKY'.

simplify-code