What problem does it solve? Teams want to contribute coding-agent delivery findings to a shared reference dataset without leaking sensitive local information such as file paths, repository names, model IDs, or fine-grained timestamps. This Skill projects the local delivery ledger into nine bounded, enum- or bucket-based fields and validates the payload before anything leaves the machine. ## Core Features & Use Cases - Privacy-first projection: Reduces each run to nine fields (kind, task class, CLI band, ISO week, terminal state, delivery state, error class, and log2 buckets for output tokens and tool calls), with a redaction self-check measured against the actual bytes. - Consent-gated sending: A dry-run preview, a literal payload review mode, and a required --send --yes confirmation ensure nothing is transmitted without explicit user approval in the current conversation. - Duplicate and live-run protection: Runs already in the ledger and runs touched within the last thirty minutes are held back, with --force available only for the duplicate case. - Use Case: After auditing recent Claude Code runs with /qruns, a user wants to contribute anonymized findings so the console's "Send your first findings" step completes, without exposing any repo names or paths. ## Quick Start Ask the assistant to run qcontrib to preview what delivery findings would be contributed, review the payload, and then confirm sending it.
