messaging
Pass
Audited by Gen Agent Trust Hub on Oct 4, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill operates on an attack surface that involves processing existing workspace documents (‘vision’, ‘customer-blueprint’, and ‘brand’) and user interview responses.
- Ingestion points: Extracts data from markdown files located in the workspace to ground business claims and define vocabulary.
- Boundary markers: Implements a strict ‘propose → confirm → lock’ multi-step workflow where the user must approve every generated section individually.
- Capability inventory: Writes and modifies multiple workspace files (‘MESSAGING.md’, ‘messaging.json’, ‘CLAUDE.md’) and updates operational data tables.
- Sanitization: Relies on mandatory human-in-the-loop verification to ensure all output is accurate and safe before being committed to the workspace.
- [COMMAND_EXECUTION]: The skill utilizes Model Context Protocol (MCP) tools (‘messaging_get_document’, ‘messaging_get_contract’, ‘messaging_query’, ‘messaging_record’) to manage its canonical documents and metadata. These represent standard, controlled tool invocations for data persistence and retrieval within the agent environment.
Audit Metadata