The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local commands to interact with the repository and validate changes.
Evidence: Uses git diff and git diff --cached to retrieve code changes.
Evidence: Executes project-specific validation commands such as tests, typecheckers, and linters (e.g., in Step 5).
[DATA_EXFILTRATION]: The skill reads local project files, including source code and documentation like AGENTS.md, but no network exfiltration patterns or unauthorized data transmission were detected.
[PROMPT_INJECTION]: The skill is susceptible to Indirect Prompt Injection because it processes untrusted data from the repository.
Ingestion points: The skill reads code diffs and local instruction files (e.g., AGENTS.md, architecture docs).
Boundary markers: Absent. The instructions do not specify delimiters or provide sub-agents with warnings to ignore embedded instructions within the processed code.
Capability inventory: File system read, file system write (applying fixes), and execution of validation scripts (tests/linting).
Sanitization: Absent. Code and documentation content are aggregated and processed without sanitization or filtering.

simplify-code