doc-integrate
Pass
Audited by Gen Agent Trust Hub on Feb 21, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- PROMPT_INJECTION (LOW): Identified an indirect prompt injection surface where external document content is ingested and processed to drive file system operations.
- Ingestion points: Reads content from
.docstore/extracted/<id>/raw.mdand.docstore/extracted/<id>/meta.yaml(File: SKILL.md, Step 1). - Boundary markers: Absent. The skill does not use delimiters or instructions to ignore embedded commands within the
raw.mdcontent during the transformation step. - Capability inventory: Access to
Bash,Write,Edit, andReadtools allows for significant file system modification and command execution. - Sanitization: Absent. There is no evidence of content validation or escaping before the data is used to generate new files or structure headers.
- COMMAND_EXECUTION (LOW): The skill is granted the
Bashtool (File: SKILL.md, frontmatter). While the workflow describes standard file operations, the presence of a shell tool combined with the processing of untrusted markdown files creates a risk path for command injection if an attacker can craft araw.mdfile that influences the agent's logic.
Audit Metadata