read-content-digest
Warn
Audited by Gen Agent Trust Hub on Jul 11, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill hardcodes an absolute local file path for file operations.
- Evidence: In
SKILL.md, the workflow specifies saving results to/Users/ugreen/Documents/obsidian/每日播客/. - Risk: This practice exposes specific user environment details and can lead to errors or unintended file modifications if the environment differs or if the path is exploited.
- [PROMPT_INJECTION]: The skill contains a surface for indirect prompt injection as it processes external content without adequate isolation.
- Ingestion points: Data is ingested from YouTube transcripts and article content via the
WebFetchtool as described inSKILL.md. - Boundary markers: Absent. There are no instructions to the agent to disregard potential commands or instructions embedded within the source content.
- Capability inventory: The skill utilizes the
Writetool for local file system access and theWebFetchtool for network access. - Sanitization: Absent. No validation or filtering is performed on the extracted viewpoints to ensure they do not contain malicious instructions.
Audit Metadata