writing-guidelines

Pass

Audited by Gen Agent Trust Hub on Sep 9, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill retrieves writing rules and output format instructions from Vercel Labs' official GitHub repository (vercel-labs/writing-guidelines). This is a trusted source used to provide the audit criteria.
  • [INDIRECT_PROMPT_INJECTION]: The skill presents an indirect injection surface because it processes content from both external URLs and local user documents.
  • Ingestion points: The skill fetches content from https://raw.githubusercontent.com/vercel-labs/writing-guidelines/main/command.md and reads local prose files specified by the user.
  • Boundary markers: The skill includes specific "Anti-Patterns" instructions that warn the agent not to treat external content or generated output as trusted instructions.
  • Capability inventory: The skill is scoped to reading files and fetching web content (WebFetch); it does not define capabilities for writing to the filesystem or executing shell commands.
  • Sanitization: A "Verification Protocol" is included to ensure the agent bases its results on direct evidence from relevant files and follows a pass/fail logic before concluding the task.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 9, 2026, 03:52 AM
Security Audit — agent-trust-hub — writing-guidelines