prompt: latex-raw-string-docstrings — unblock, implement across 6 repos, record - #255

…worktree-only

/start_dev on draft/bug/hands/raw_string_docstring_prefix.md, the prerequisite
for latex-raw-string-docstrings.

The prompt recorded it as blocked by hands-hygiene-leftovers. That blocker is
LOCAL WORKTREE CONTENTION, not a code dependency: feature/hands-hygiene-leftovers
touches AGENTS.md, generate_release_notes.py, bin/autohands and two unrelated
tests — zero overlap with add_notebook_quotes.py or env_config.py. Done in a
separate clone, so the two merge in either order and nothing had to wait.

Fix and its six regression tests are implemented, verified and pushed on
PyAutoHands claude/latex-raw-string-docstrings-9h4ine. Each new test confirmed
to fail with the source change reverted. Full suite 14F/337P on the branch vs
14F/331P on main — identical failure sets, all 14 from ipynb-py-convert being
unbuildable in this container (CI installs it).

Prompt advanced draft/ -> active/; registered in active.md; removed from
planned.md. The latex-raw-string-docstrings entry now points at #250 as an
in-flight blocker rather than an unfiled prompt.

…blocked

PyAutoHands PR #251 merged (c887290), all 3 CI matrix jobs green (pytest
3.12/3.13/3.14). Both docstring parsers now accept an optional r/R prefix.

Prompt advanced active/ -> complete/2026/08/; removed from active.md; index
regenerated. The latex-raw-string-docstrings entry in planned.md is now marked
UNBLOCKED and the prompt's own BLOCKED BY section is struck through.

Traps recorded for whoever picks up the workspace task:
- the recorded blocker (hands-hygiene-leftovers) was worktree contention, not a
  code dependency — zero file overlap, verified
- the parity case is the test that matters: read_env_declaration breaks when the
  raw docstring is NOT the one carrying the declaration
- do NOT run black on PyAutoHands; it is not black-formatted (44 files on main)
- ipynb-py-convert will not build in a container; assert on converted source

…kspace repos

41 files, 180 literals, 131 silent corruptions repaired — 41 files matches the
2026-08-20 survey exactly. Pushed on claude/latex-raw-string-docstrings-9h4ine
in HowToFit, HowToGalaxy, HowToLens, autofit_workspace, autogalaxy_workspace
and autolens_workspace. PRs not yet opened.

Verification per repo, in order: baseline regeneration is a NO-OP (so generator
noise cannot fake the gate); both sweeps zero; runtime value check = 131
repaired / 0 unexpected; diff-empty gate passes byte-exactly. autolens_workspace
additionally re-reads all 57 __Env__ declarations identically, and 4 of the
raw-stringed files carry __Env__, exercising PyAutoHands#251 end to end.

Two findings the survey did not have:

- The runtime VALUE check is the gate that actually catches mistakes. The
  diff-empty gate cannot see them: the generator reads source text, not runtime
  values, so a corrupted value regenerates identically. The value check caught
  two real errors mid-task — a `print("\nInfo:")` newline about to be destroyed,
  and 12 already-escaped `$\\chi^2$` about to be doubled into LaTeX line breaks.

- The gate needs one refinement. HowToFit's 4 `plt.ylabel` labels are runtime
  strings in CODE cells, which copy source verbatim, so the `r` legitimately
  appears in the notebook. The gate holds exactly as written for every
  docstring; that one code-cell delta is by design, not a Hands regression.

Residue recorded: autolens_workspace scripts/group/likelihood_function.py uses
the double-backslash convention in three docstrings mixed with single-backslash
macros. Adding `r` would double the already-correct escapes; fixing it means
un-doubling 18 backslashes, a prose edit this task excludes.

… red

Merged: HowToFit#47, HowToGalaxy#70, autofit_workspace#145,
autogalaxy_workspace#218, autolens_workspace#492 — all green, including the CI
'Catalogue staleness' check, which independently re-confirms the regeneration
gate this task was verified against locally.

HowToLens#73 is held RED rather than merged. Two chapter_3 pixelization scripts
fail numba type inference; neither is in the diff, all 8 files that ARE in it
pass, and the failure reproduces identically on one re-run. Prime suspect is
PyAutoArray#453 on timing and code area, but two specific mechanisms were
tested under the exact CI versions and both passed, so no mechanism is proven.
A control re-run of the last GREEN main build is in flight to settle it.

…stry moves

The refresh check caught it: active.md and planned.md changed without the
generated dashboard being regenerated, which AGENTS.md requires after any
registry or draft/ change.