recording-to-help-article
Pass
Audited by Gen Agent Trust Hub on Sep 26, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTIONPROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests and processes external recording data to generate documentation, which represents a potential attack surface for data-embedded instructions.
- Ingestion points: Source recordings and recording transcripts (SKILL.md, Workflow Step 2).
- Boundary markers: The instructions do not define specific delimiters to separate processed data from internal agent logic.
- Capability inventory: The skill uses Clueso MCP tools to create articles, attach screenshots, and update project content (SKILL.md, Workflow Steps 5-9).
- Sanitization: There is no mention of sanitizing or filtering the content extracted from recordings.
- [COMMAND_EXECUTION]: The skill instructs the user to manually execute commands to configure the environment (e.g.,
claude mcp add ...). These commands are necessary for the skill's operation and are presented for user review and confirmation. - [EXTERNAL_DOWNLOADS]: The workflow involves connecting to an MCP server hosted at
https://connect.clueso.io/mcp. This domain is controlled by the vendor and is a legitimate part of the skill's infrastructure. - [PROMPT_INJECTION]: Instructions include logic to suppress conversational output regarding workspace selection when only a single workspace is available. This is a design choice for user experience and does not constitute malicious concealment of operations.
Audit Metadata