arkcli-helper

Pass

Audited by Gen Agent Trust Hub on Sep 7, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes the arkcli binary and npx -y @byted-supabase/cli@latest install to configure agents and install necessary vendor tools.
  • [INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect injection because it processes user prompts through a routing mechanism (routing.v1) to determine which tool (e.g., doubao_search, datapro) to invoke.
  • Ingestion points: Intercepts user inputs at UserPromptSubmit and PreToolUse gates in files like SKILL.md and references/arkcli-helper.md.
  • Boundary markers: Uses a structured JSON schema for Route Plans and requires validation against a per-turn ID.
  • Capability inventory: Performs file writes to agent configuration directories (e.g., ~/.claude.json, ~/.codex/config.toml), injects environment variables into .env files, and invokes system commands through the arkcli utility.
  • Sanitization: Enforces a strict mapping between classified intents and specific authorized providers.
  • [EXTERNAL_DOWNLOADS]: Fetches vendor-specific packages via npx and provides instructions for the manual installation of the Evolve skill from the vendor-controlled domain volces.com.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 7, 2026, 04:04 PM
Security Audit — agent-trust-hub — arkcli-helper