feat(labs): make every lab a self-contained, skip-friendly starting point

Each lab now stands on its own; no hard dependency on prior labs.
- App-based labs get a canonical tasks-app snapshot in lab/start/ (three
  baselines: v0 add/list/done; v1 +count; v2 +count/delete), assigned by where
  each module sits in the command timeline. Modules with a purpose-built app
  (M10 trap, M13 planted bug, M21) snapshot their own app; planted devices kept.
- Self-contained labs (M15/17/18/19/22/23/24/25/27, which operate on their own
  lab files) get a preamble pointing at modules/NN/lab/.
- Every module + capstone gets a "Starting point (skip-friendly)" preamble:
  copy the snapshot, git init -b main, commit -> clean status, then start.

Lets a learner skip around or recover: copy start/, commit, go. All snapshots
run; tools/check.sh passes; no em-dashes.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01TfzV5QvtPDz8LJS3Pu5VLT

modules/27-evals/README.md

@@ -212,6 +212,15 @@ That's the durable skill. Models are weather. The eval set is the thermometer yo
 
 ## Hands-on lab
 
+
+> **Starting point (this lab is skip-friendly).** This lab is self-contained and does not depend
+> on the earlier labs. Its files live in `modules/27-evals/lab/`. Copy them into a working folder
+> and make a first commit so you start clean:
+>
+> ```bash
+> cp -r ~/ai-workflow-course/modules/27-evals/lab ~/ai-workflow-course/27-evals-lab
+> cd ~/ai-workflow-course/27-evals-lab && git init -b main && git add -A && git commit -m "start: module 27"
+> ```
 **Lab language:** Python + shell. You'll run a tiny eval harness, point an agent at a task, and run
 a regression eval across a "model swap."