The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted natural language data from external files to drive its code and test generation logic.
Ingestion points: proposal.md, features/*.feature, design.md, and tasks.md located in the beat/changes/<name>/ directory.
Boundary markers: The skill does not implement specific boundary markers or instruction-bypassing warnings when reading artifact content.
Capability inventory: Includes file writing for code and tests, shell command execution for test runners (npx, pnpm, behave), and invocation of other agent 'superpowers'.
Sanitization: No explicit sanitization or filtering of the content from the processed artifacts is described.
[COMMAND_EXECUTION]: The skill invokes several command-line tools to execute tests and perform file comparisons.
Evidence: Shell execution of npx cucumber-js, pnpm test, behave, and diff is integrated into the workflow.
[DYNAMIC_EXECUTION]: The skill generates and modifies implementation and test code at runtime.
Evidence: The skill loop in SKILL.md (Step 6) details the generation of automated tests and implementation code based on input scenarios and technical decisions.

apply