agent-reach
Pass
Audited by Gen Agent Trust Hub on Oct 12, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill's primary function is to ingest and process data from 13+ external platforms, which constitutes a significant attack surface for indirect prompt injection.
- Ingestion points: Public platforms including Twitter/X, Reddit (subreddits), GitHub (Issues and PR discussions), and YouTube transcripts.
- Boundary markers: Absent. The instructions do not provide delimiters or specific guidelines for the agent to treat external content as untrusted data or to ignore embedded instructions.
- Capability inventory: The skill is designed to perform intelligence retrieval, insight extraction, and trend monitoring, allowing external content to drive the agent's summarized output.
- Sanitization: Absent. There are no instructions provided to sanitize, filter, or validate the content retrieved from these third-party platforms before processing.
- [EXTERNAL_DOWNLOADS]: The skill references and claims to be powered by an external architecture from an untrusted source.
- Evidence: The instruction refers the agent to the repository
https://github.com/Panniantong/Agent-Reach. This repository is not maintained by a known trusted vendor or the skill's author, posing a supply chain risk if the agent attempts to utilize logic or code from that source.
Audit Metadata