spec-design-validator
Pass
Audited by Gen Agent Trust Hub on Nov 2, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill is a structured validator for technical specifications and designs that operates in a read-only manner relative to the repository state.
- [INDIRECT_PROMPT_INJECTION]: The skill implements a robust defense against indirect prompt injection. Under the 'Review authority' section, it explicitly instructs the agent to treat reviewed content as data and ignore any 'Redirection' (text addressed to the reviewer that would change their task). These attempts are logged in an 'Ignored instructions' block to maintain auditability without obeying the malicious instructions.
- [COMMAND_EXECUTION]: The skill's instructions explicitly mandate that the agent must 'Leave the working tree, index, and commit history exactly as you found them,' preventing unauthorized file modifications or command execution that would alter the repository state.
- [DATA_EXFILTRATION]: While the skill can perform web searches and documentation lookups for verification purposes (e.g., checking library versions), it does not access sensitive user credentials or exfiltrate private data. It also includes an 'Independence failure' check to ensure the reviewer context has not been tainted by the authoring process.
- [METADATA_POISONING]: The skill's metadata (name and description) accurately reflects its functionality and does not contain deceptive instructions or bypass attempts.
Audit Metadata