business-data-context
Pass
Audited by Gen Agent Trust Hub on Oct 5, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONPERSISTENCE
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill possesses a potential attack surface for indirect prompt injection because it processes data from the live database and backend repository. Ingestion points: Database query results, source code from the backend repository, and the local HTML data map. Boundary markers: Instructions mandate a 'proposal' workflow where the agent must wait for human confirmation before applying changes. Capability inventory: Database access and scoped file writing to the references directory. Sanitization: Mandatory verification of facts against the live database and explicit user approval required for all updates.
- [COMMAND_EXECUTION]: The skill uses the command 'bun run context:map' to parse its data map. This is a local script execution within the user's project environment and is part of the skill's core functionality.
- [PERSISTENCE]: The skill updates its own reference files to persist knowledge about the system's data structure. This behavior is clearly defined and restricted to a specific write scope in the skill's metadata to prevent unauthorized file system modifications.
Audit Metadata