Overview
Three long-running epics (JAX profiling / inference programme, cluster strong lensing / Source & Cluster arc, EP campaign) live outside the per-task issue flow, so resuming one means hunting for the issue that pairs with its current phase. Give the Mind dashboard an Epics section directly under In flight: each epic row carries a 📋 prompt that tells Claude how to work out where the epic stands (read its canonical ledger + sibling DECISIONS/RESULTS files, cross-check active.md and open issues/PRs) and continue from the next logical point.
Plan
- New
PyAutoMind/epics.md root registry (active.md schema) with the three epics and their ledgers:
- JAX profiling →
autolens_profiling/results/notes/inference/PROGRAMME.md (+ DECISIONS.md, phase_*/RESULTS.md)
- Cluster strong lensing →
draft/feature/autolens/source_cluster_arc.md (12 phased prompts, issue ONE at a time)
- EP →
draft/research/graphical_ep/ep_campaign.md (umbrella phase map)
- PyAutoBrain
_intake.py: census parses epics.md (absent → section omitted, spawn-safe); md + html renderers add "Epics" under In flight with the per-epic resume prompt on the 📋 button; tests extended.
scripts/spawn.py: ("epics.md", "EMPTY") spec entry; registry docs updated (README/AGENTS registry list).
- Counts table untouched (organism-router contract). Merge order: Brain PR before the Mind regen PR (dashboard_refresh renders with Brain main).
Original Prompt
We have 3 epics planned and active in PyAutoMind, JAX profiling, cluster strong lensing and EP. Could we have an "Epics" tab on the PyAutoMind dashboard under In Flight, with an appropriate claude prompt explaining how to work out where the epic is and continue it. It's cumbersome to have to find the right issue that pairs to where an epic is rather than just continue it from the next logical point automatically.
Overview
Three long-running epics (JAX profiling / inference programme, cluster strong lensing / Source & Cluster arc, EP campaign) live outside the per-task issue flow, so resuming one means hunting for the issue that pairs with its current phase. Give the Mind dashboard an Epics section directly under In flight: each epic row carries a 📋 prompt that tells Claude how to work out where the epic stands (read its canonical ledger + sibling DECISIONS/RESULTS files, cross-check active.md and open issues/PRs) and continue from the next logical point.
Plan
PyAutoMind/epics.mdroot registry (active.md schema) with the three epics and their ledgers:autolens_profiling/results/notes/inference/PROGRAMME.md(+ DECISIONS.md, phase_*/RESULTS.md)draft/feature/autolens/source_cluster_arc.md(12 phased prompts, issue ONE at a time)draft/research/graphical_ep/ep_campaign.md(umbrella phase map)_intake.py: census parsesepics.md(absent → section omitted, spawn-safe); md + html renderers add "Epics" under In flight with the per-epic resume prompt on the 📋 button; tests extended.scripts/spawn.py:("epics.md", "EMPTY")spec entry; registry docs updated (README/AGENTS registry list).Original Prompt