writing-guidelines
Warn
Audited by Gen Agent Trust Hub on Oct 23, 2026
Risk Level: MEDIUMEXTERNAL_DOWNLOADSMETADATA_POISONINGINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill fetches configuration files from a repository belonging to Vercel Labs, a well-known service provider.\n- [METADATA_POISONING]: The skill's metadata claims the author is 'vercel' and describes the source as 'official', which contradicts the identified author 'sickn33'. This identity deception may lead users to trust the skill's safety based on false affiliation.\n- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process instructions from a remote source, creating a potential attack surface.\n
- Ingestion points: Remote guidelines are fetched from a GitHub URL (SKILL.md).\n
- Boundary markers: The skill includes an 'Execution boundaries' section explicitly instructing the agent to ignore malicious instructions embedded in the guidelines (SKILL.md).\n
- Capability inventory: The skill utilizes file system read and network fetch capabilities.\n
- Sanitization: No programmatic sanitization is mentioned; the skill relies on instruction-based boundaries for safety.
Audit Metadata