prompt: file issue for workspace-agent (#116) + route → start_library

Author: Jammy2211

active.md

@@ -1,6 +1,14 @@
 # Active Tasks
 
 
+## workspace-agent
+- issue: https://github.com/PyAutoLabs/PyAutoBrain/issues/116
+- status: library-dev
+- worktree: ~/Code/PyAutoLabs-wt/workspace-agent
+- autonomy: supervised
+- repos:
+  - PyAutoBrain: feature/workspace-agent
+
 ## pixelized-gradient-experiment
 - issue: https://github.com/PyAutoLabs/autolens_workspace_developer/issues/100
 - status: in-progress — A100 pipeline WORKING. Gradient question SETTLED (yes). Search question OPEN. Nautilus baseline job 330379 running on RAL (submitted 2026-07-15, 4h limit).

docs/autolens/i_want_to_use_the_autolens_assistant.md

@@ -0,0 +1,43 @@
+# I want to use the autolens_assistant to perform Expectation Propagation
+
+Type: docs
+Target: PyAutoLens
+Repos:
+- PyAutoLens
+- autolens_assistant
+- autolens_workspace
+Difficulty: too-large
+Autonomy: supervised
+Priority: high
+Status: formalised
+
+I want to use the autolens_assistant to perform Expectation Propagation (EP) analysis of a Cosmology science case.
+
+The science case is inferring the Hubble constant from time delay lensed quasars, and a example package already
+containing EP scripts and runs, simulators and other key things is at:
+
+/mnt/c/Users/Jammy/Science/concr/scritps/cosmology and /mnt/c/Users/Jammy/Science/concr/simulator/cosmology.py
+
+This is actually quite a mature project, but its a git old so probably has some API drift (e.g. old PyAutoLens)
+and generally needs to be brushed up.
+
+There are also EP and other examples in autolens_workspace/scripts/guides/modeling/advanced
+
+Can you therefore make a new science project using the autolens_assistant, which uses the main scripts in thee projects
+project to perform the EP fit but also allows for the one by one / graphical fits. However, for now, I want to descope
+down from the Hubble coonstant for now, and have the only goal be to simulate N strong lens using the power-law + shear
+mass model, where the power-law slopes are drawn from a hierarchical distribution. The example scripts should
+then be built around recovering the slopes acurate but, more important the mean and scatter of the slope hierarchical
+parameter.
+
+The goal are:
+
+1) Show that we can do this for large samples without EP, using JAX gradient samples (SPECIFY).
+
+2) Show that we can do this with EP, ideally showing it recovers the same values and errors.
+
+3) We are going to scale up to large lens samples, so make sure all of this can run on RAL and its HPC via the HPC link.
+
+4) To test all the graphical EP diagnostics and result analysis code, especially tyhe recent EP updates we did last week.
+
+<!-- formalised by the Intake (Conception) Agent on 2026-07-16 from file:/tmp/claude-1000/-home-jammy-Code-PyAutoLabs/32708468-5918-4dbc-a763-583805364341/scratchpad/intake_ep_cosmology.md -->