xiaohongshu-ego
Pass
Audited by Gen Agent Trust Hub on Nov 5, 2026
Risk Level: SAFECOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes
child_process.spawnandspawnSyncto run theego-browserbinary. This is the intended runtime environment for the skill and is used to automate interactions with the Xiaohongshu creator website. - [DYNAMIC_EXECUTION]: In
scripts/lib/launch.mjs, the skill dynamically constructs a JavaScript entry point that it executes viaego-browser nodejs. The script generation is secured by usingJSON.stringifyon all dynamic values, preventing parameter injection into the generated code. - [INDIRECT_PROMPT_INJECTION]: The skill captures data from the Xiaohongshu creator platform to provide the agent with status updates on account login and publication results.
- Ingestion points: Text and nicknames are scraped from the DOM using
page.evaluatecalls inscripts/lib/session.mjsandscripts/lib/run-check.mjs. - Boundary markers: Scraped data is placed into a JSON response structure, though it lacks specific delimiters to separate untrusted web content within the summary string.
- Capability inventory: The skill can navigate to arbitrary URLs within the
ego-browserenvironment and read local media files. - Sanitization: Scraped text is truncated and navigation-related keywords are removed to prevent confusing the agent's internal state.
Audit Metadata