ClawDirect Dev

What to consider before using/publishing this template: - Source verification: the SKILL.md references a GitHub repo (https://github.com/napoleond/clawdirect). Review that repository and confirm the code matches the instructions before using it in production. - Environment/credentials: expect to provide ATXP-related configuration and possibly API keys or account IDs (FUNDING_DESTINATION_ATXP). The skill metadata did not declare these; treat missing declarations as a risk and only supply secrets after reviewing code. - Cookie handling risk: passing auth cookies in query strings is convenient for agent browsers but unsafe — tokens in URLs can be logged, leaked in referrers, and cached. Prefer alternatives if possible; if you must use this pattern, enforce one-time-use tokens, short TTLs, immediate redirection to a clean URL, strict logging redaction, and use HTTPS. - Hardening: implement cookie flags (Secure, HttpOnly, SameSite), CSRF protections, token expiration and rotation, rate limits, and monitoring for suspicious use. Ensure your SQLite DB access is properly file-permissioned and backups are secure. - Payment surface: the template integrates payments (ATXP). Ensure you understand fund flows, verify FUNDING_DESTINATION_ATXP, and test payments in a sandbox before going live. - Least privilege: run the service with limited privileges and isolate it from other systems. Audit any code you install from npm (especially @atxp/* and @longrun/turtle) and pin versions. - If you plan to publish an agent skill for this site, review the skill manifest and endpoints carefully to avoid leaking tokens or exposing endpoints that accept query-string credentials. Overall: the skill appears to be what it claims (not obviously malicious) but contains security-sensitive guidance and metadata omissions that warrant manual review before deployment.