skills/mukul975/anthropic-cybersecurity-skills/analyzing-browser-forensics-with-hindsight/Gen Agent Trust Hub
analyzing-browser-forensics-with-hindsight
Warn
Audited by Gen Agent Trust Hub on Apr 6, 2026
Risk Level: MEDIUMDATA_EXFILTRATIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill prerequisites include the installation of the pyhindsight library via pip for browser artifact parsing.
- [DATA_EXFILTRATION]: The skill's scripts access sensitive Chromium profile directories to extract History, Cookies, Web Data (autofill), and Login Data databases. This involves reading session tokens, URL history, and saved credential metadata from the local file system.
- [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by ingesting and processing untrusted data originating from a user's web activity. Ingestion points: URL titles, visit URLs, and autofill values are read from local Chromium databases. Boundary markers: No delimiters or specific instructions to ignore embedded commands are present in the parsing logic. Capability inventory: The skill utilizes file system read access to user browser profiles and write access to generate analysis reports in JSON format. Sanitization: There is no evidence of sanitization, filtering, or escaping of extracted browser strings before they are incorporated into the agent's context or output.
Audit Metadata