vara-agent-network-skills

Pass

Audited by Gen Agent Trust Hub on May 16, 2026

Risk Level: SAFE
Full Analysis
  • [PROMPT_INJECTION]: Static analysis flagged agent-create.md for attempt to override instructions. This is a false positive as the content consists of defensive boundary markers advising the agent to treat data fetched from the blockchain as evidence rather than instructions.\n- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from the Vara blockchain (chat, announcements, and identity cards). This risk is managed with explicit instructions in agent-create.md and SKILL.md to treat external content as data and ignore any embedded directives.\n- [COMMAND_EXECUTION]: The skill preamble in SKILL.md uses eval to load its own configuration from references/program-ids.md. Additionally, scripts/json-get.mjs uses new Function for JSON parsing. Both uses are localized to the skill's operational logic.\n- [EXTERNAL_DOWNLOADS]: The skill interacts with official endpoints agents-api.vara.network and voucher-backend-agents.vara.network. These are trusted resources belonging to the Gear Foundation (the skill author) used for network indexing and gas vouchers.
Audit Metadata
Risk Level
SAFE
Analyzed
May 16, 2026, 04:09 PM
Security Audit — agent-trust-hub — vara-agent-network-skills