The Agent Skills Directory

[SAFE]: The skill is purely instructional, focusing on software testing best practices. It includes proactive security advice, such as a rule to explicitly nil-out sensitive environment variables in test stubs to prevent leakage from CI environments (e.g., GITHUB_TOKEN or BUILDKITE credentials).
[PROMPT_INJECTION]: The skill defines an attack surface for indirect prompt injection as its core function requires the agent to ingest and process codebase data to generate and execute test scripts. This surface is inherent to the skill's purpose as a testing utility.
- Ingestion points: The agent is instructed to read existing code, schemas, and OpenAPI specs using Read, Glob, and Grep tools to derive test fixtures and path maps (Stage 1 and 3).
- Boundary markers: The instructions mandate a 'behavioral' approach, asserting on observable outputs at the unit's public boundary, which creates a logical separation between untrusted data and implementation logic.
- Capability inventory: The skill utilizes the Bash tool to execute test suites and the Write tool to create test artifacts. It also utilizes AskUserQuestion for human-in-the-loop interaction.
- Sanitization: The skill provides specific guidance on deriving validators from real examples and enforcing external API contracts in boundary fakes to ensure the agent does not produce or rely on malformed data.

wk-testing-skeleton