cost-optimize

Pass

Audited by Gen Agent Trust Hub on Jun 13, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: Downloads the platform CLI tool via npx @claude-flow/cli@latest from the NPM registry. As NPM is a well-known service and the CLI is a core platform component, this is a safe operational dependency.
  • [COMMAND_EXECUTION]: Executes a local outcome-tracking script (plugins/ruflo-cost-tracker/scripts/outcome.mjs) to update model routing based on success or failure. The skill specifies that the script uses safe execution methods (spawnSync) to prevent argument injection.
  • [PROMPT_INJECTION]: Surfaces a potential indirect prompt injection vulnerability (Category 8).
  • Ingestion points: Usage data retrieved from the cost-tracking namespace.
  • Boundary markers: None implemented in the instructions.
  • Capability inventory: Local shell execution via Bash and pattern storage via memory_store.
  • Sanitization: The instruction set notes the use of safe quoting in the downstream shell script.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 13, 2026, 01:23 PM
Security Audit — agent-trust-hub — cost-optimize