docs: update stale admin_jammy/prompt/ refs to PyAutoPrompt/ - #1

`scripts/pyauto_audit.sh` defines a `pyauto-audit` shell function (sourced
from ~/.bashrc next to pyauto_status.sh) that scans ~/Code/PyAutoLabs/
for state the dashboard can't show:

  1. Top-level directories with no .git (skip "." and "z_" prefixes
     for hidden / personal-staging dirs).
  2. Stashes older than PYAUTO_AUDIT_STASH_DAYS (default 14 days).
  3. Local-only branches with no upstream and last commit older than
     PYAUTO_AUDIT_BRANCH_DAYS (default 30 days).

Three sections, plain text, sections suppressed when empty, single
"clean" message when all clear. Always exits 0 — informational, the
user reads and decides. Run on demand. PYAUTO_AUDIT_ROOT,
PYAUTO_AUDIT_STASH_DAYS, PYAUTO_AUDIT_BRANCH_DAYS env-var overrides.

Re-scoped from autoprompt 06 (the heavyweight monthly cron audit
spec). The dashboard's `b` flag (shipped in #10) covers prompt 06's
findings #1 + #2; prompts 02 + 07 cover #4 + #7. Only the genuinely
structural checks remain.

Also deletes autoprompt/04_source_of_truth_rule.md — already-rejected
during this sweep. The pyauto-status BEHIND counts and prompt 03's
history-rewrite guard cover the failure modes that 04 was trying to
address; keeping a rejected prompt in autoprompt/ is just noise.

Co-authored-by: Jammy2211 <JNightingale2211@gmail.com>
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

…olens_profiling repo)

…ofiling#2)

Move 2_modeling_cluster to Shipped (cluster-modeling-v2, autolens_workspace#175,
2026-05-18). Sequence the cluster/2..5 follow-up prompts (scaling relation,
test workspace, likelihood function walkthrough, profiling) as Outstanding,
each linked from the z_feature. Drop scaling-relation members from Deferred
(now Outstanding #1); keep start_here.py rewrite, lens/source CSV API, and
aplt plotter promotion as Deferred.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

… phase 2 continues

…ship sign-off; phase 4 queued (#1)

…e 1 shipped)

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

… content = Phase 2-3

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

…o-end

The record claimed the PAT question needed a real run. It did not: the secret
inventory is API-readable, and PAT_PYAUTOLABS was absent from PyAutoMind
entirely, so Monday would have failed the guard. After adding it, the path was
proved by inducing drift and dispatching — the self-heal opened PR #1 in 24s
with the correct diff, merged as 7f8576f, --check back to 0.

…ete/

Both entries read as history, not in-flight work: release-drive-2026-08-03
was marked SHIPPED/superseded, and release-drive-2026-08-07 shipped all five
libraries to PyPI at 2026.8.7.1 and was verified against pypi.org.

Written via `lifecycle.py record --date 2026-08-07`, which folds each section
into complete/2026/08/, prunes it from active.md and refreshes complete/index.md
in one step. Content is carried over verbatim -- the trap catalogues are the
point of the rich record (rehearsal #1 discarded on the mid-build merge, the
ingest trap, the pre_build untracked-WIP hazard, the post-publish PyPI
index-propagation race). Only the `repos-none-claimed` line is dropped from
each: it documents an active.md formatting constraint that means nothing once
the entry leaves that file.

active.md is left holding the two genuinely live entries (pix-prodigy-gpu-compat,
heart-green-validation-ingest). lifecycle check + index --check both OK.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01FTzN9AEF8DY2jTX2qxVHRk