feat: pyauto-status dirty file listing + split MOD/UNTR columns

[Jammy2211]

Overview

Follow-up to #2 / PR #3. The pyauto-status dashboard currently shows a single DIRTY count per repo, with no detail about what is dirty. After surfacing real noise on first use (autolens_base_project 14 dirty, admin_jammy 4, etc.), it's clear we need (a) the actual file list per repo and (b) a way to track whether untracked junk is building up — i.e. when .gitignore needs to be extended.

This is a small enhancement, scoped only to PyAutoPrompt/scripts/pyauto_status.sh.

Plan

• Cache git status --porcelain output once per repo (instead of running git status twice — once for the count, again would be needed for the listing)
• Replace the single DIRTY column with two: MOD (tracked-modified files) and UNTR (untracked). The flag glyph * keeps firing if either is non-zero
• After the main table, append a Dirty files: section listing the porcelain output for each repo with dirty content, indented under the repo name. Skip repos with nothing dirty
• No auto-categorization of "looks generated" — the ?? vs M prefixes from porcelain output already tell the user what's untracked
• Day-to-day signal: if UNTR climbs across runs for a repo, the .gitignore is missing patterns

Detailed implementation plan

Affected Repositories

• PyAutoPrompt (only)

Work Classification

Library (treated as such for the standard /start_library → worktree → /ship_library flow).

Branch Survey

Repository	Current Branch	Dirty?
./PyAutoPrompt	main	clean

Suggested branch: feature/dashboard-dirty-listing
Worktree root: ~/Code/PyAutoLabs-wt/dashboard-dirty-listing/ (created later by /start_library)

Implementation Steps

1. Edit PyAutoPrompt/scripts/pyauto_status.sh:

   • In the per-repo data-collection loop, capture porcelain="$(git -C "$repo" status --porcelain 2>/dev/null)" into a bash associative array keyed by repo name (declare -A repo_porcelain).
   • Compute mod and untr separately from the cached porcelain output: mod is lines NOT starting with ??, untr is lines starting with ??. Use grep -c against the cached string.
   • The flag column logic stays: * if (mod + untr) > 0.

2. Update the printf format string and column headers:

   • Old: '%-32s %-30s %-36s %6s %5s %5s %s\n' with headers REPO BRANCH UPSTREAM BEHIND AHEAD DIRTY FLAGS.
   • New: '%-32s %-30s %-36s %6s %5s %4s %4s %s\n' with headers REPO BRANCH UPSTREAM BEHIND AHEAD MOD UNTR FLAGS.
   • Adjust the dashed separator row to match.

3. After the main row loop, emit the dirty-file listing:

   • Iterate repos again. For each repo with non-empty ${repo_porcelain[$name]}, on the first hit print a blank line and Dirty files: header. Then print $name: followed by the porcelain content indented by 4 spaces (sed 's/^/ /').
   • No cap on file count — load-bearing signal, not noise. Can revisit if a repo overflows in practice.

Key Files

• PyAutoPrompt/scripts/pyauto_status.sh — single-file change.

Acceptance Verification

• time pyauto-status from a fresh shell still completes in <10s (no new git invocations).
• A repo with 0 dirty shows MOD=0 UNTR=0, no * flag, and is omitted from the listing.
• A repo with both modified-tracked and untracked files shows non-zero in both columns, and both appear in the listing with the correct M / ?? prefixes.
• bash PyAutoPrompt/scripts/status.sh --repos still produces the new format (no regression on the delegation path).

Out of scope

• Auto-categorizing untracked files as "likely generated" by extension — fragile, user-eye is faster.
• Tracking dirty counts across runs (would need persistent state).
• Color/highlight — kept plain text.

Original Prompt

This issue does not have a prompt-file source — it's a direct follow-up enhancement to PR #3 surfaced during day-1 use of pyauto-status.