The Agent Skills Directory

[DATA_EXFILTRATION]: The skill accesses and extracts highly sensitive information from browser profile directories, including decrypted passwords, session cookies, and personal identity information (PII).
Evidence: parser.get_credentials(), parser.get_cookies(), and FormAnalyzer.get_credit_cards() functions in SKILL.md.
Evidence: The skill explicitly searches for sensitive strings like password, api_key, and token within the browser cache.
[DATA_EXFILTRATION]: The skill facilitates connection to the well-known VirusTotal service for threat intelligence checks on downloaded files.
Evidence: analyzer.check_virustotal(api_key=VT_API_KEY) in SKILL.md allows checking file metadata against an external database.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from web artifacts which could potentially contain malicious content designed to influence the agent's behavior during analysis.
Ingestion points: The skill ingests history records, bookmarks, extension manifests, and cached web content from the provided browser profile paths in SKILL.md.
Boundary markers: The documentation does not specify the use of delimiters or instructions to ignore embedded commands within the analyzed artifacts.
Capability inventory: The skill performs extensive file system reads, report generation (generate_report), and external network calls (check_virustotal).
Sanitization: There is no evidence of sanitization or escaping of the ingested browser data before it is included in analysis reports or processed by the agent.

browser-forensics