light-writing-for-agents
Pass
Audited by Gen Agent Trust Hub on Oct 16, 2026
Risk Level: SAFENO_CODEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill consists entirely of instructional Markdown files and metadata. No scripts, binaries, or executable commands are present within the package.
- [NO_CODE]: The content is purely educational, providing guidelines for writing documents for agents. It does not implement any programmatic logic or tool invocations.
- [INDIRECT_PROMPT_INJECTION]: The skill provides instructions for the agent to create and edit files based on user prompts. The ingestion points include documents like AGENTS.md and CLAUDE.md. The instructions lack specific boundary markers or sanitization guidelines for external content, and the agent typically uses its internal file-system capabilities to execute these modifications.
Audit Metadata