The Agent Skills Directory

[SAFE]: No malicious prompt injection patterns were detected. The skill uses strong instructional language such as "The Iron Law" and "CRITICAL" to enforce the Test-Driven Development (TDD) methodology (e.g., writing failing tests before production code), which is standard instructional behavior for this domain and not an attempt to bypass agent safety guidelines.
[SAFE]: No data exfiltration or unauthorized credential exposure was found. Code examples in mocking.md correctly demonstrate the use of environment variables (process.env.STRIPE_KEY) for secret management, which is a safe practice.
[SAFE]: No obfuscation techniques such as Base64 encoding, zero-width characters, or homoglyph attacks were detected in any of the skill files.
[SAFE]: The skill does not perform any external downloads, remote script execution, or unverifiable package installations. It refers to standard project files like CONTEXT.md and .feature files and delegates tasks to another internal skill (/mattpocock:tdd).
[SAFE]: No privilege escalation or persistence mechanisms (such as modifications to shell profiles or cron jobs) are present.
[SAFE]: The skill demonstrates a standard software development workflow with a minimal attack surface. While it ingests external data (user requirements and CONTEXT.md) and utilizes the AskUserQuestion tool, these are used appropriately within the context of the BDD lifecycle.
Ingestion points: Reads CONTEXT.md and user-provided feature requirements.
Boundary markers: Not explicitly defined in the prompt templates.
Capability inventory: Uses the AskUserQuestion tool to confirm test seams; no direct system command execution or network exfiltration capabilities are present in the skill code.
Sanitization: Not explicitly defined, but the intended output is Gherkin scenarios and test code, which carries low risk of downstream injection when handled by an AI agent.

bdd