refactor: apply review feedback (round 1)

Objective: Code review identified 6 issues across CI workflow,
shell scripts, SKILL.md references, and .gitignore.

Approach: Accept 6 items, reject 4 items with technical reasoning.
- CI: remove nonexistent --fix flag from error message, fix
  unreachable exit-code logic with set +e
- Shell: add executable bit (100755) to detect-env.sh, hybrid-health.sh
- SKILL.md: add eval-metrics.md to reference table, fix quick-eval.py
  usage from --input/--reference flags to positional args
- .gitignore: remove duplicate __pycache__ entry (already covered
  by **/__pycache__/ on line 53)

Evidence: git ls-tree confirms 100755 for .sh files.
grep confirms no remaining --fix references in CI.
SKILL.md quick-eval.py example matches argparse positional interface.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Author: hyunhee-jo

.github/workflows/skill-drift-check.yml

@@ -25,10 +25,12 @@ jobs:
 
       - name: Check skill drift
         run: |
+          set +e
           python skills/odl-pdf/scripts/sync-skill-refs.py
-          if [ $? -ne 0 ]; then
+          EXIT_CODE=$?
+          if [ $EXIT_CODE -ne 0 ]; then
             echo ""
             echo "Drift detected: skill references are out of sync with options.json."
-            echo "Run 'python skills/odl-pdf/scripts/sync-skill-refs.py --fix' locally to update them."
+            echo "Update skills/odl-pdf/references/options-matrix.md to match options.json."
             exit 1
           fi

.gitignore

@@ -75,5 +75,3 @@ logs/
 # Configuration files
 .claude/settings.local.json
 .claude/plans/
-
-skills/odl-pdf/scripts/__pycache__/

skills/odl-pdf/SKILL.md

@@ -363,9 +363,7 @@ When extraction quality is inadequate. Start with measurement, then escalate.
 Run the quick evaluation script against your output:
 
 ```bash
-python skills/odl-pdf/scripts/quick-eval.py \
-  --input output/document.json \
-  --reference ground-truth.json
+python skills/odl-pdf/scripts/quick-eval.py output/document.md ground-truth.md
 ```
 
 Or run the full benchmark to get NID, TEDS, and MHS scores:
@@ -686,6 +684,7 @@ Load these files progressively — only when entering the relevant topic. Do not
 | `references/options-matrix.md` | User needs detailed option documentation, defaults, or interactions |
 | `references/hybrid-guide.md` | User needs hybrid server setup, server-side flags, or remote deployment |
 | `references/format-guide.md` | User needs output format comparison, format-specific behavior, or format selection |
+| `references/eval-metrics.md` | User needs detailed metric definitions (NID, TEDS, MHS), benchmark scores, or diagnostic steps by metric |
 | `scripts/detect-env.sh` | Phase 1 environment detection — run at session start |
 | `scripts/quick-eval.py` | Phase 4 quality measurement — run when diagnosing extraction quality |
 | `evals/` | Benchmark baselines and regression thresholds |