skills/base/skills/build-on-base/Gen Agent Trust Hub

build-on-base

Pass

Audited by Gen Agent Trust Hub on Sep 18, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [PROMPT_INJECTION]: The skill includes behavioral directives in references/agents/register.md (Phase 4) that attempt to establish permanent rules for the agent, specifically requiring that all future transactions include builder code attribution.
  • [PROMPT_INJECTION]: Detailed interactive workflows in references/migrations/farcaster-miniapp-to-app.md define a wizard-like quiz flow that overrides the agent's default processing to enforce a specific diagnostic and path-selection routine.
  • [COMMAND_EXECUTION]: The skill provides several local scripts (scripts/register.sh, scripts/analyze_project.py, scripts/validate_conversion.py) that perform network calls and local file analysis. These are intended to be executed by the agent to automate registration and migration tasks.
  • [INDIRECT_PROMPT_INJECTION]: The analysis scripts ingest project files and source code to identify migration needs, creating a surface where instructions embedded in external code could influence the agent's actions during the conversion process.
  • [EXTERNAL_DOWNLOADS]: The scripts/register.sh script makes POST requests to https://api.base.dev/v1/agents/builder-codes. This communication is directed to the vendor's official API for core functionality.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 18, 2026, 10:27 AM
Security Audit — agent-trust-hub — build-on-base