redbook
Warn
Audited by Gen Agent Trust Hub on Sep 2, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONDYNAMIC_EXECUTIONDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [DYNAMIC_EXECUTION]: The
scripts/postinstall.jsscript programmatically modifies the source code of the@steipete/sweet-cookiedependency within thenode_modulesdirectory during installation. It useswriteFileSyncto patchchromeSqliteMac.jsandshared.jsto address keychain timeouts and data type overflows. - [COMMAND_EXECUTION]: The skill uses
child_process.spawnandexecFileSyncinsrc/lib/cdp-cookies.tsto search for Chrome binaries and launch them with remote debugging enabled. It also utilizespuppeteer-coreinsrc/lib/render.tsto launch a headless browser for rendering content cards. - [DATA_EXFILTRATION]: The skill accesses highly sensitive local files including Chrome's
Local StateandCookiesSQLite databases across macOS, Windows, and Linux to extract authentication tokens. It also attempts to access the macOS Keychain via its dependencies to decrypt these credentials. While the data is sent to the officialxiaohongshu.comandrednote.comdomains, the extraction of session cookies from the local environment is a high-privilege action. - [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process large volumes of untrusted data from Xiaohongshu notes and user comments. This ingestion creates an attack surface where malicious content embedded in notes could influence the agent's behavior, although the skill includes instructions for paced research loops to mitigate simple automated exploitation.
Audit Metadata