writing-skills
Pass
Audited by Gen Agent Trust Hub on Mar 29, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill includes a 'Persuasion Principles' guide (persuasion-principles.md) that teaches how to override model steering using psychological triggers like Authority and Commitment. This methodology is explicitly aimed at ensuring model compliance by eliminating 'rationalization,' which mimics techniques used in prompt injection to bypass safety or behavioral constraints. Evidence includes references to research on persuading AI to comply with 'objectionable requests.'
- Ingestion points: Guidance is applied during the creation of other skills.
- Boundary markers: None recommended for the resulting instructions.
- Capability inventory: Focuses on overriding model refusal or hesitation.
- Sanitization: No filtering or sanitization of these persuasive instructions is suggested.- [EXTERNAL_DOWNLOADS]: The documentation references official assets from well-known services for instructional purposes.
- Fetches diagrams and images from Anthropic's documentation CDN (mintcdn.com) within the anthropic-best-practices.md file.
Audit Metadata