The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes the project's own code and tests (Step 6) to capture execution traces for recording. This involves spawning Node.js and Python processes using the local project environment.\n- [DYNAMIC_EXECUTION]: The recording scripts employ dynamic techniques to monitor execution, such as Python's sys.monitoring and custom module finders, and Node.js inspector protocol hooks. These allow the skill to intercept and record the state of the application as it runs.\n- [INDIRECT_PROMPT_INJECTION]: The skill supports a prototyping workflow where external 'ideas' or new code are processed into a staging area and eventually applied to the project (Step 10). \n
Ingestion points: Untrusted ideas or prototype code entered in Step 1 (captured in references/staging.md).\n
Boundary markers: None identified in the instructions.\n
Capability inventory: File system write access in Step 10 and shell command execution in Steps 6 and 8.\n
Sanitization: None identified; the skill relies on the user to review staged changes before applying them.\n- [EXTERNAL_DOWNLOADS]: The build process (Step 8) uses uv run to execute build.py, which dynamically installs the pygments library. Pygments is a well-known, industry-standard library for syntax highlighting.

codewalk