timeframe-research
Pass
Audited by Gen Agent Trust Hub on Sep 30, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest data from untrusted external sources (GitHub, Hugging Face, Reddit, Hacker News, arXiv) and summarize it into the local workspace.
- Ingestion points: Web search results and community platform content processed by subagents as described in Step 4.
- Boundary markers: The skill uses year and month headings to structure the output but lacks specific escape characters or instruction-ignoring delimiters for the summarized content.
- Capability inventory: The skill can launch subagents and write content directly to specified files in the workspace (Steps 3 and 4).
- Sanitization: Instructions in Step 5 and 6 mandate 'neutral language' and 'no hype,' which provides a degree of protection against accidental instruction following.
- [DYNAMIC_EXECUTION]: The skill implements a recursive subagent architecture, dynamically generating prompts and launching year and month-specific agents as defined in Step 3 and 4 of SKILL.md.
- [COMMAND_EXECUTION]: The README.md suggests installation via 'npx skills add av/skills', which executes a CLI tool from the npm registry.
- [EXTERNAL_DOWNLOADS]: The skill references and downloads components from the av/skills repository on GitHub.
Audit Metadata