dbs-content-system
Pass
Audited by Gen Agent Trust Hub on Nov 11, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process a large volume of untrusted local text assets (documents, tweets, etc.) from the user's workspace. This creates a surface for indirect prompt injection if those assets contain malicious instructions intended to influence the agent's extraction or assembly logic.\n
- Ingestion points: The skill reads raw content from the
01-原始素材区/完整副本/directory.\n - Boundary markers: The skill enforces a strict output schema (YAML frontmatter and Markdown headings) for extracted content units, which provides some structural delimitation.\n
- Capability inventory: The skill has the capability to read/write files and execute local Node.js utility scripts.\n
- Sanitization: No specific semantic sanitization is applied to the content before it is presented to the agent for structural analysis.\n- [COMMAND_EXECUTION]: The skill's operational model relies on the agent executing a suite of provided Node.js scripts (e.g.,
extract-sample-units.js,rebuild-processing-ledger.js) to manage the content project. These scripts perform file system operations and data processing within the workspace. The scripts are transparently provided within the skill's own directory structure.
Audit Metadata