The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the Bash tool and explicitly instructions the agent to perform command execution for verification and script permission management (e.g., chmod +x). Additionally, the generated scripts themselves (Node.js, Python, Rust, Go) are intended to execute shell commands at runtime.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it reads untrusted data from project manifest files (e.g., package.json scripts, .claude/runner-config.json) and interpolates these values directly into templates that are then written as executable scripts.
Ingestion points: Reads package.json, .claude/runner-config.json, pyproject.toml, and various lock files.
Boundary markers: None identified; values are interpolated into template strings (e.g., ${TEST_SCRIPT}, ${LINT_GLOBS_JSON}) without escaping or boundary markers.
Capability inventory: The skill has Write and Bash capabilities, allowing it to create executable files and run them.
Sanitization: There is no evidence of sanitization or validation of the values read from project files before they are written into the runner scripts.
[EXTERNAL_DOWNLOADS]: The skill references templates but these are local to the skill package (references/templates.md). No external network downloads were detected.

generate-runner