The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill utilizes the mcp__speed-run__generate_and_write_files tool to create source code based on hosted LLM output. This code is subsequently executed through the implementation of tasks and the Test-Driven Development (TDD) cycle.
[COMMAND_EXECUTION]: Several shell commands are executed to manage the development environment, including git worktree operations (add, remove) and directory comparisons using diff to detect implementation variations.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and act upon content from a design.md file, which serves as the primary instruction source for parallel runners.
Ingestion points: The runner agents read docs/plans/<feature>/design.md at the start of their workflow.
Boundary markers: There are no explicit boundary markers or safety instructions provided to the sub-agents to ignore potentially malicious instructions embedded within the design document.
Capability inventory: The runners have access to file-writing tools, shell command execution via git, and code generation/execution tools.
Sanitization: The skill does not perform validation or sanitization on the design document before it is processed by the runners.

showdown