vault-maintain
Pass
Audited by Gen Agent Trust Hub on Apr 18, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: Orchestrates local maintenance workflows by calling specialized sub-skills including vault-ingest, vault-lint, vault-tracker, and vault-concepts. These executions are restricted to specified operational modes (report, apply-safe, apply) for vault management.
- [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface due to its data ingestion capabilities. 1. Ingestion points: Reads unprocessed data from raw/sources/ and curated pages (SKILL.md). 2. Boundary markers: No explicit delimiters or instructions to ignore instructions within ingested data are defined. 3. Capability inventory: Includes file moves (archive), file writes (log.md, concept pages), and local command execution (SKILL.md). 4. Sanitization: No sanitization or validation of the content being ingested or summarized is mentioned.
Audit Metadata