The Agent Skills Directory

[COMMAND_EXECUTION]: The helper script scripts/lint_runner.py is designed to execute shell commands from the target repository, such as npm run lint or local binaries like eslint and tsc located in node_modules/.bin.
Evidence: The script uses subprocess.run() with argument lists to execute these commands, avoiding direct shell interpretation where possible.
Mitigation: The skill instructions explicitly warn the agent to manually inspect package scripts before execution, noting that they can run arbitrary project code.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted files from external repositories, creating a surface for potential exploitation if the repository contains malicious configuration or scripts.
Ingestion points: scripts/lint_runner.py parses package.json for script definitions and scripts/type_coverage.py reads source code for analysis.
Boundary markers: There are no technical delimiters separating data from instructions within the scripts, but the SKILL.md file provides operational boundaries through its mandatory procedures.
Capability inventory: The skill possesses the capability to read project files and execute shell commands based on external data.
Sanitization: The implementation employs secure practices, including subprocess.run without shell=True, bounded file reads (max 1MB and 30 files) in the type coverage tool, and symlink avoidance during directory traversal.

lint-and-validate