byted-kickart-viral-replicator

Pass

Audited by Gen Agent Trust Hub on Oct 8, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill utilizes the kickart-open-cli tool through npm exec to perform operational tasks such as OAuth authentication (auth_login, auth_complete), task submission (submit_viral_task_v2), and status querying (query_task_v2).- [EXTERNAL_DOWNLOADS]: The skill dynamically fetches the official @volcengine/kickart-open-mcp package from the well-known NPM registry (registry.npmjs.org) to ensure it uses the latest vendor-approved version of the management tools.- [DYNAMIC_EXECUTION]: The skill implements dynamic runtime execution by invoking the CLI tool and parsing its structured JSON output to manage the lifecycle of marketing video generation.- [INDIRECT_PROMPT_INJECTION]: The skill manages an attack surface for indirect injection by ingesting and processing untrusted external inputs such as product URLs and creative prompts.\n
  • Ingestion points: Creative prompts (--prompt), product links (--product-url), and reference video metadata in SKILL.md and 爆款裂变指南.md.\n
  • Boundary markers: The skill explicitly requires a '成片信息' (Final Video Info) confirmation phase where the user must review all parameters before submission.\n
  • Capability inventory: The skill has the capability to execute shell commands via npm exec and read/write task metadata to the local filesystem.\n
  • Sanitization: The instructions mandate strict desensitization of URLs and the suppression of authentication tokens in command outputs to prevent data leakage.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 8, 2026, 09:13 AM