meta-ad-manager
Pass
Audited by Gen Agent Trust Hub on Oct 27, 2026
Risk Level: SAFEDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [DYNAMIC_EXECUTION]: The skill is instructed to run a local validation script (
contract/validate.py) using the Python interpreter to verify the structure and contents of the ad management directory. This is a controlled execution of a script bundled with the skill for quality assurance. - [INDIRECT_PROMPT_INJECTION]: The skill ingests and processes data from various files in the
ads/directory, which can include content generated by external tools (e.g., Meta reports, Shopify summaries) or user input. - Ingestion points: Files including
ads/README.md,ads/brand.md, and campaign-specific files (state.md,strategy.md,decisions.md). - Boundary markers: The skill uses structured Markdown headings and YAML frontmatter blocks to delimit data fields.
- Capability inventory: The skill can perform file writes, call Meta API tools (launch, pause, read), and route tasks to other installed skills.
- Sanitization: The skill employs a dedicated Python validator (
validate.py) to enforce the structural integrity and schema of the documents it processes.
Audit Metadata