docs: add /health and /checkpoint to README (v0.16.2.2)

[ignsm]

Summary

Follow-up to #951. Both /health and /checkpoint have been working skills for a while — health/SKILL.md and checkpoint/SKILL.md live on disk, have clean frontmatter with trigger phrases, and ./setup already registers them into ~/.claude/skills/ on every install. But they were completely absent from the README in all three user-facing places: the install command skill list (line 49), the CLAUDE.md example skill list (line 398), and the big skill table (around line 186). New users had no way to discover them.

This PR closes that gap. All three places now mention both skills.

Stacking note

This is a stacked PR on top of #951 (fix/ngrok-external-in-node-bundle). #951 aligned the two README skill-list blocks and added /pair-agent, /plan-devex-review, /devex-review. This PR builds on that cleanup to add the remaining two skills in one coherent pass.

Until #951 merges, this PR's diff will show all 4 commits from #951 plus the 4 new commits here. Once #951 merges to main, GitHub will auto-rebase this PR and only the 4 new commits will remain. No manual rebase needed.

If you want to review just this PR's delta: check commits 12c7bbf1, 0f7e5222, 89fe0a4c, fd6d4f6e.

Bisection

Four clean commits, each independently revertable:

1. 12c7bbf1 — docs: add /health and /checkpoint rows to README skill table
   • Adds one row for each skill to the big sprint skill table. /health after /benchmark (both measurement/dashboard tools). /checkpoint after /learn (both session/state persistence tools).
   • Specialist role choices: /health → "Quality Engineer" (sits alongside "Performance Engineer" for /benchmark). /checkpoint → "Session Bookmark" (avoids collision with /learn's "Memory" role since they're adjacent).
   • Voice matches the existing table: one-line role, two-to-three-sentence description emphasizing what the user can DO and the concrete trigger phrase.
2. 0f7e5222 — docs: add /health and /checkpoint to both README skill-list blocks
   • Adds both skills to the install command block (line 49) and the CLAUDE.md example block (line 398). Same placement as the table: /health after /benchmark, /checkpoint after /learn.
   • Both blocks now list the same 36 skills in the same order (verified via extracted-list diff).
3. 89fe0a4c — docs: mark /checkpoint + /health README TODO as done
   • Removes the ## Documentation section from TODOS.md (it existed only to hold this one entry) and moves the entry to ## Completed with a v0.16.2.2 marker.
4. fd6d4f6e — chore: bump version and changelog (v0.16.2.2) (VERSION + CHANGELOG.md)

Test Coverage

No new code paths — this is a pure documentation change across 3 files (README.md, TODOS.md, CHANGELOG.md) and a VERSION bump. No application logic, no build scripts, no runtime behavior changes.

Both skills were manually verified to exist and register correctly on the /Users/ignat/GitHub/gstack checkout before writing the table rows:

• health/SKILL.md lines 1-13 (frontmatter with description, allowed-tools, trigger phrases)
• checkpoint/SKILL.md lines 1-13 (frontmatter with description, allowed-tools, trigger phrases)
• ./setup output on the upgrade session that surfaced this gap confirmed both skills land in ~/.claude/skills/checkpoint/SKILL.md and ~/.claude/skills/health/SKILL.md as working top-level slash commands.

Pre-Landing Review

No issues. 5 substantive line edits across README.md (2 new table rows, 2 skill-list string updates) + 14 lines added/removed in TODOS.md + 3 lines in VERSION/CHANGELOG. No attack surface, no runtime effect, no regression risk.

Voice calibration

Both descriptions were written to match the existing skill table voice:

• Lead with what the user can DO, not with the implementation.
• Include the concrete trigger phrase in quotes so the reader knows how to activate the skill.
• Avoid AI vocabulary (no "leverage", "robust", "comprehensive"). No em dashes in the new text (matches CLAUDE.md writing rules).
• Pick a specialist role that doesn't collide with adjacent table entries. /learn is already "Memory", so /checkpoint is "Session Bookmark". /benchmark is already "Performance Engineer", so /health is "Quality Engineer".

Test plan

• grep "^## \[" CHANGELOG.md shows contiguous version sequence (verified — 0.16.2.2, 0.16.2.1, 0.16.2.0, 0.16.1.0, ...)
• Both README skill-list blocks contain the same 36 skills (verified via extracted-list diff: install block and CLAUDE.md example block both produce identical sorted lists)
• /health and /checkpoint appear in all three places: install block (line 49), CLAUDE.md example (line 398-403), big skill table (after /benchmark and after /learn respectively)
• Both skills resolve to real SKILL.md files: health/SKILL.md exists, checkpoint/SKILL.md exists, ./setup links both into ~/.claude/skills/ (confirmed from prior setup run on the upgrade session)

🤖 Generated with Claude Code

[ignsm]

Closing this along with its parent #951. The README work from this PR (adding /health and /checkpoint to both skill-list blocks and the big skill table) has been combined with the skill-list alignment from #951 into a single clean README-only PR: #953.

The split into a parent + stacked follow-up didn't make sense once #951 turned out to duplicate open build-fix PRs #924 and #947. #953 is the correct final shape: one branch from fresh main, three bisected commits, no overlap with any build-fix work in flight.