The Agent Skills Directory

[COMMAND_EXECUTION]: The skill requires the agent to execute standard test runner commands (such as pytest, vitest, jest, or go test) to verify software behavior. These operations are restricted to the project's local environment and are necessary for the TDD process.
[PROMPT_INJECTION]: The skill processes untrusted external data, creating an attack surface for indirect prompt injection.
Ingestion points: The agent reads user-provided tickets, source code files, and stdout/stderr output from test runners as part of the loop in SKILL.md.
Boundary markers: The skill uses structured updates (Stage / Found / Next / Needs user) and a predefined summary template for outputs, which helps maintain context but does not explicitly wrap input data in safety delimiters.
Capability inventory: The skill allows execution of arbitrary shell commands for testing purposes, as documented in references/test-commands.md.
Sanitization: No explicit sanitization or filtering of test output or code content is specified before the agent processes it.
[SAFE]: The skill follows established software engineering best practices. It includes a robust evaluation suite (evals/) and clear instructions for the agent to defer to the user in ambiguous or high-risk situations, such as changes involving authentication or data migration.

tdd-loop