qbr-plan
Pass
Audited by Gen Agent Trust Hub on Oct 13, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted external data sources, creating a potential surface for indirect prompt injection. \n
- Ingestion points: Processes configuration files and business data from
~/.claude-marketing/brands/(profiles, guidelines, SOPs) as well as user-supplied campaign metrics and client satisfaction signals. \n - Capability inventory: The skill is restricted to performance analysis, strategic recommendation generation, and report structuring. No evidence was found of network operations, shell command execution, or access to sensitive system credentials. \n
- Boundary markers: No explicit delimiters or 'ignore instructions' warnings were identified to prevent the agent from potentially obeying instructions embedded within the ingested campaign data. \n
- Sanitization: The instructions do not define validation or filtering mechanisms for the data provided by the user or extracted from local files.
Audit Metadata