What problem does it solve?
Sharing a coding-agent session for harness feedback normally risks leaking secrets, personal data, or customer information, and manually assembling a complete, auditable report is error-prone. This Skill builds a sanitized, consent-gated support bundle so maintainers can diagnose agent behavior without exposing private data.
Core Features & Use Cases
- Complete session export handling: Accepts a native JSON session export or retrieves an explicitly identified active Codex thread via a bundled read-only exporter, verifying first and latest turn coverage.
- Local sanitization pipeline: The
prepare-share-bundle.mjs script redacts credentials, tokens, emails, IPs, phone numbers, private keys, and absolute paths, then produces a sanitized session.json, a ZIP of session-generated files, a manifest with SHA-256 hashes, and a privacy report.
- Hard consent gate and atomic publication: Nothing is published until the user reviews the exact artifacts and types an explicit acknowledgement; publication then creates a deduplicated public branch and an upstream harness-feedback issue, with rollback if issue creation fails.
- Use Case: After a coding agent hits a rate limit or produces a broken feature, you say "share this session" to file a complete, privacy-reviewed diagnostic report with the Open Mercato maintainers.
Quick Start
Ask the agent to share this session under a public slug like "share this session as login-flow-bug" and follow the review and consent prompts.