writing-skills
Pass
Audited by Gen Agent Trust Hub on Sep 17, 2026
Risk Level: SAFEPROMPT_INJECTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The instructions incorporate specific psychological persuasion principles such as Authority and Commitment to enforce strict compliance with defined rules. The skill uses absolute and imperative language such as 'YOU MUST STOP', 'No exceptions', and 'Delete means delete' to command the agent to override its own reasoning or decision-making process when deviations occur.
- [INDIRECT_PROMPT_INJECTION]: The skill establishes a testing methodology that utilizes 'pressure scenarios' with subagents to identify and close 'loopholes' in their behavioral compliance. * Ingestion points: Stress-test scenarios and rationalization tables used to verify skill efficacy in references/testing-skills.md and references/tdd-mapping.md. * Boundary markers: None identified between testing prompts and skill instructions. * Capability inventory: Modification and deletion of files in skill directories (~/.claude/skills). * Sanitization: Absent; the agent is encouraged to adopt and iterate on persuasive techniques.
- [PROMPT_INJECTION]: The 'Iron Law' section in the skill documentation mandates destructive actions if specific process requirements are not met, which could lead to the unintended loss of data if the agent applies these rules zealously to project files during the authoring process.
Audit Metadata