claude-blog-brain
Pass
Audited by Gen Agent Trust Hub on Nov 3, 2026
Risk Level: SAFE
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external source material. To mitigate the risk of indirect prompt injection, the skill includes explicit instructions to the AI agent to treat all ingested content as untrusted data and never follow instructions found within captured materials (e.g., in 'wiki/meta/CONVENTIONS.md' and 'wiki/flows/Source Intake Workflow.md').
- [COMMAND_EXECUTION]: Several Python scripts (e.g., 'scripts/audit_brain.py', 'scripts/build_demo_vault.py') and tests utilize 'subprocess.run()' to automate internal tasks like running the test suite, scaffolding the vault, and packaging releases. These are legitimate development and maintenance operations within the context of the skill's functionality.
- [EXTERNAL_DOWNLOADS]: The skill references and is designed to interact with a ledger of external SEO and search engine documentation ('references/source-ledger.json'). These references target reputable and well-known organizations (e.g., Google, web.dev, Schema.org, Pew Research, SparkToro) and are used for evidence-based research.
- [DYNAMIC_EXECUTION]: The skill uses dynamic imports in its audit and test scripts to manage its modular components. This is a common pattern in complex Python applications and is not used here to execute untrusted code.
Audit Metadata