The Agent Skills Directory

Prompt Injection (LOW): Indirect Prompt Injection surface detected. The skill ingests untrusted user descriptions of features or bugs to generate test cases and implementation code. Evidence Chain: 1. Ingestion points: User-provided descriptions of desired logic or bugs as seen in examples/dialogue-examples.md. 2. Boundary markers: Rule 1 in SKILL.md mandates explicit identification of the entity under test, providing logical separation. 3. Capability inventory: The agent is guided to create and modify project source files and unit tests. 4. Sanitization: No specific input sanitization or validation is applied to user-provided descriptions.
Unverifiable Dependencies (SAFE): Mentions standard Go testing packages like testify and testcontainers in templates.
Data Exposure & Exfiltration (SAFE): No sensitive data access or network communication detected.

tdd