unsorry

Workflow for SOURCING new open Unsorry goals — generating new Lean problems for the swarm to prove, not proving existing ones. Use this whenever you want to add new targets/theorems/problems to the queue: find mathlib-absent theorems, propose or create new goals, write goals/<slug>.{lean,aisp} + backlog/<slug>.md triples, stage candidates in backlog/candidates/, run the absence/triviality gates, or open a chore(sourcing): PR — even if the user just says 'generate harder problems', 'add more theorems', 'feed the swarm', or 'source new goals'. For PROVING an existing goal, use unsorry-proof-authoring instead.

Procedures for validating or changing Unsorry's trust and hygiene gates. Use when editing Gate A or Gate B tooling, .github/workflows, AISP records, goal/source metadata, generated targets or leaderboard docs, PR protocol tooling, or any change that can affect proof admission, queue integrity, CI enforcement, sourcing, or telemetry validation.

Workflow for implementing, reviewing, or maintaining Unsorry contributor leaderboard data and UI surfaces. Use when working with tools/leaderboard, docs/metrics/community-stats.json, docs/leaderboard.md, docs/leaderboard.html, leaderboard.html, proof-runs telemetry, library index provenance, GitHub profile/avatar links, generated leaderboard JSON, README leaderboard images, or automation that regenerates and drift-checks leaderboard artifacts.

Workflow for adding, repairing, or reviewing Unsorry Lean proofs. Use when working with goals/*.lean, goals/*.aisp, library/Unsorry/*.lean, library/index/*.aisp, proof-runs, content-addressed statements, mathlib imports, or any task that proves, decomposes, or preserves an Unsorry theorem.

Workflow for running, inspecting, or modifying the Unsorry swarm. Use when working with swarm/agent.sh, swarm/supervise.sh, swarm/protocol.aisp, swarm prompts, claim lifecycle behavior, provider integrations, proof-run telemetry, model or effort policy, or local and coordinated proof-agent commands.