user-guide-documents
Pass
Audited by Gen Agent Trust Hub on Aug 12, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTIONDATA_EXFILTRATION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted external data from sources such as Slack messages, engineering notes, and support tickets to generate its output. Ingestion points: Slack messages, engineering notes, and ticket details described in SKILL.md. Boundary markers: Absent; the skill relies on instructional rules rather than structural delimiters. Capability inventory: The agent can write to a local file (MEMORY.md) and interact with the Notion API via tools. Sanitization: The skill provides clear instructions to redact internal identifiers, ticket IDs, and technical jargon.
- [DATA_EXPOSURE_AND_EXFILTRATION]: The instruction to capture screenshots of browsers or applications represents a potential risk for the accidental exposure of sensitive environment data if the agent fails to properly apply the specified redaction steps.
- [PERSISTENCE_MECHANISMS]: The skill uses a local MEMORY.md file to store and retrieve Notion page identifiers and URLs, providing a method for maintaining state across different sessions.
Audit Metadata