The Agent Skills Directory

[SAFE]: The skill utilizes standard, well-known testing frameworks and utilities including Vitest, Jest, pytest, and cargo-nextest. It refers to official documentation from trusted organizations such as Anthropic and established technical authorities like NIST and JUnit. The skill promotes secure practices such as data anonymization and secret management (e.g., using .env files). No patterns of malicious exfiltration, persistence, or privilege escalation were detected.
[INDIRECT_PROMPT_INJECTION]: The skill analyzes external project code and test outputs to perform its functions, which exposes it to indirect prompt injection if malicious instructions are embedded in code comments or data processed by the agent. The skill includes specific internal policies to mitigate this, such as requiring assertions to be derived from independent specifications rather than the implementation code and enforcing test isolation.
Ingestion points: Reads source code files, dependency lockfiles, and test execution logs from the local workspace (SKILL.md, unit-testing.md).
Boundary markers: The 'Core Contract' in SKILL.md requires auditing expected-value provenance to prevent circular verification and mandates that every test be isolated from shared state.
Capability inventory: The skill can modify source code by adding or repairing test files and executes shell commands to run language-specific test suites (integration-testing.md, multi-language-testing.md).
Sanitization: Employs industry-standard mocking and stubbing libraries (MSW, WireMock, Mockito) to isolate external dependencies and control data flow (integration-testing.md).

radar