preserve-implementation-intent
Pass
Audited by Gen Agent Trust Hub on Sep 23, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONNO_CODE
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes external and untrusted data sources to generate technical artifacts, which could contain malicious instructions.
- Ingestion points: The skill reads conversation history, exploration findings, existing assets (proposals, designs, specs, tasks), and referenced code or documentation in
SKILL.mdunder the 'Before Drafting' section. - Boundary markers: The skill includes instructions to distinguish agreed requirements from observed behavior and unverified assumptions, providing a conceptual boundary, though not a cryptographic or rigid one.
- Capability inventory: The skill itself does not invoke tools for network operations, subprocess execution, or direct file system manipulation outside of standard artifact generation.
- Sanitization: There are no explicit instructions for sanitizing or escaping the content of the ingested data.
- [NO_CODE]: The skill consists entirely of markdown-based instructions for the agent's reasoning process and does not include any scripts, binaries, or command-line execution patterns.
Audit Metadata