accelint-skill-prose

Pass

Audited by Gen Agent Trust Hub on Aug 21, 2026

Risk Level: SAFE
Full Analysis
  • [COMMAND_EXECUTION]: The README provides standard installation instructions using npx and pnpm pointing to the vendor's official repository gohypergiant/agent-skills. These are routine package management operations conducted through trusted developer tools.
  • [INDIRECT_PROMPT_INJECTION]: As a prose auditor, the skill processes external prompt artifacts which could theoretically contain instructions designed to influence agent behavior. However, the skill implements rigorous boundary markers (e.g., 'Untouchables' list), mandatory self-checks, and a fixed output template to ensure that the content being audited does not override the agent's core operating parameters or safety instructions.
  • [DATA_EXPOSURE]: The skill explicitly defines sensitive technical references, such as file paths, identifiers, and field names, as 'Untouchables' that must be preserved exactly. This serves as a defensive mechanism to prevent the accidental modification or exposure of sensitive system configuration data during the prose optimization process.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 21, 2026, 03:22 PM
Security Audit — agent-trust-hub — accelint-skill-prose