The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill's primary function is to ingest user-described technical conflicts and generate functional code demos to resolve them, creating an injection surface.
Ingestion points: User-provided descriptions of the "divergence" and conflicting positions in Phase 1 of the pipeline (SKILL.md).
Boundary markers: No specific delimiters or "ignore embedded instructions" warnings are prescribed to isolate the user-provided conflict text from the code generation instructions.
Capability inventory: The skill enables shell command execution via project task runners (Rule 2) and file-system writes to create the demos.
Sanitization: The instructions do not include steps to sanitize or validate the user-provided logic or UI descriptions before they are used to generate executable scripts.
[DYNAMIC_EXECUTION]: The skill is designed to generate and execute code at runtime to verify logic models or UI variations.
Evidence: Phase 2 and Rule 2 (SKILL.md) detail the creation of "disposable code" that is run via the project's task runner.
Evidence: The LOGIC mode in REFERENCE.md explicitly instructs building portable modules and interactive TUI shells that are executed locally.
[COMMAND_EXECUTION]: The skill requires the ability to invoke shell commands to run the produced benchmarks and prototypes.
Evidence: Rule 2 (SKILL.md) mandates a "one command to run" success contract, typically utilizing tools like pnpm, make, or just.

have-a-try