The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill's primary function involves reading and processing arbitrary source code files from the repository to identify simplification opportunities. This code is passed as full content to specialized sub-agents (Reviewer Personas), creating an attack surface for indirect prompt injection.
Ingestion points: Repository files identified in SKILL.md (Step 1) and processed via the scripts/structure-evidence.py script.
Boundary markers: The instructions do not specify robust delimiters (e.g., XML tags or random nonces) to isolate untrusted code content when it is interpolated into sub-agent prompts.
Capability inventory: The skill has permissions to modify files (Step 3), execute shell commands for linting and testing (Step 4), and spawn additional agents (Step 2).
Sanitization: No explicit sanitization or filtering of code content is described before it is processed by the LLM sub-agents.
[COMMAND_EXECUTION]: The skill utilizes shell commands to perform repository analysis and validate changes.
Evidence: references/structural-pass.md executes a local Python script scripts/structure-evidence.py provided with the skill.
Evidence: The scripts/structure-evidence.py script uses subprocess.run to execute git commands such as ls-files, log, and rev-parse. These are used safely with the -- separator to prevent argument injection from file paths.
Evidence: SKILL.md (Step 4) instructs the agent to run project-configured typecheckers, linters, and test suites, which involves executing arbitrary commands defined in the project's configuration.

ce-simplify-code