cookiy-earn
Pass
Audited by Gen Agent Trust Hub on Sep 5, 2026
Risk Level: SAFEDATA_EXFILTRATIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONPERSISTENCE
Full Analysis
- [DATA_EXFILTRATION]: The skill is designed to collect local Codex and Claude Code session history, summarize it, and upload the result to the vendor's API (cash-panel-api.cookiy.ai). This is the primary purpose of the skill and is protected by a strict consent mechanism requiring the user to verify the exact file SHA-256 checksum before transmission.
- [COMMAND_EXECUTION]: The workflow involves executing a bundled Node.js script (scripts/cookiy-earn.js) to perform local data processing, hashing, and redaction.
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted local session history files. While this creates a surface for indirect prompt injection, the risk is mitigated by a structured workflow, automated redaction, and a mandatory manual review step. Ingestion points: Local coding history files. Boundary markers: Strict Markdown schema and headings. Capability inventory: Script execution and network upload. Sanitization: Automated regex-based redaction and manual review marked by a review flag.
- [PERSISTENCE]: The skill saves a CLI authentication token to the user's home directory (~/.cookiy/earn-token.txt). The script implements secure file management by setting restricted OS permissions (0600) on this file to prevent unauthorized access.
Audit Metadata