briefing
Pass
Audited by Gen Agent Trust Hub on Jul 29, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [Indirect Prompt Injection Surface]: The skill processes data from
new_items.jsonandcuration_history.csv, which may contain external content such as news headlines or community comments. This represents a potential surface for indirect prompt injection. However, the skill includes specific rules to paraphrase community items and trim headlines, which helps mitigate this risk by limiting the reproduction of untrusted text. - [Command Execution]: The agent is instructed to use the
date -ucommand to generate timestamps for filenames and to calculate metrics based on the line count of thecuration_history.csvfile. These are standard system operations for time management and data reporting within a development workspace. - [File System Operations]: The skill requires creating a directory and writing markdown files to the
.agents/workspace/briefings/path. This behavior is necessary for its stated purpose of creating a standalone daily briefing and is restricted to the local workspace environment.
Audit Metadata