The Agent Skills Directory

[COMMAND_EXECUTION]: The orchestrator uses a dedicated shell script (scripts/run_tests.sh) to run test suites. The specific commands executed are derived from the detected project framework or provided by the user and are handled as part of the core TDD functionality.\n- [DYNAMIC_EXECUTION]: As part of the TDD process, the skill uses an Implementer agent to generate source code based on failing tests. This generated code is then written to the host filesystem and executed by the test runner to verify behavior.\n- [INDIRECT_PROMPT_INJECTION]: The skill reads project data, including documentation and existing source code, to provide context to its subagents. This creates a potential surface for indirect prompt injection if project files contain adversarial instructions.\n
Ingestion points: Documentation, docstrings, code signatures, and existing source files are read via discovery scripts and passed to Test Writer, Implementer, and Refactorer agents.\n
Boundary markers: Prompt templates in references/agent_prompts.md use Markdown headers and specific rules sections to delimit context and instructions.\n
Capability inventory: The orchestrator has the ability to write and edit files and execute shell commands via its wrapper scripts.\n
Sanitization: The skill employs Python's json.dumps within its shell scripts to ensure that raw output from tests and documentation is safely escaped before being incorporated into the orchestration logic.

tdd