ads-copy
Pass
Audited by Gen Agent Trust Hub on Jun 29, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill uses the
WebFetchtool to crawl business websites and extract information for ad copy generation, which introduces untrusted external data into the agent's context. - [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection due to its processing of untrusted website content without specific sanitization or boundary markers. This risk is managed via mandatory human-in-the-loop confirmation steps before any modifications are committed to the ad account. 1. Ingestion points: Website content retrieved from user-provided or discovered URLs via
WebFetchinSKILL.md. 2. Boundary markers: Absent in the instructions provided for processing external content. 3. Capability inventory: Includes ad creation (createAd), asset updates (updateAdAssets), and campaign management (enableAd,pauseAd) through themcp__adsagenttoolset. 4. Sanitization: No explicit sanitization or filtering of the retrieved website content is described. - [COMMAND_EXECUTION]: The skill invokes various MCP tools from the
mcp__adsagentnamespace to list campaigns, retrieve keyword reports, and perform ad performance analysis as part of its core functionality.
Audit Metadata