Fifth and final Glimmer template. Pipeline is proven. The new risk is specific to this engine: bridge severance timing. With Sam starting at 65 and decaying ~5/turn when ignored, she severs around week 6 — earlier than the "turn 8-10" target in the spec. Step 3 determines whether this needs calibration before student use.
LABOR: Claude Code
CONTEXT REQUIRED: glimmer-care-ethics/generate.js, glimmer-care-ethics/index.html, glimmer-care-ethics/example-game.json, Zebonastic project root
Deploy The Network game to the Zebonastic site.
1. Install canvas if not already present:
npm list canvas || npm install canvas
2. Copy game files:
glimmer-care-ethics/index.html
glimmer-care-ethics/generate.js
glimmer-care-ethics/example-game.json → rename to game.json
3. Validate:
node glimmer-care-ethics/generate.js \
glimmer-care-ethics \
glimmer-care-ethics/index.html \
--validate
Expected: clean pass (6 relationships, 2 bridges, 12 turns).
Any errors → stop and report.
4. Deploy:
node glimmer-care-ethics/generate.js \
glimmer-care-ethics \
glimmer-care-ethics/index.html \
--nextjs .
5. Verify:
ls -la public/games/the-network/
ls -la public/games/the-network/assets/
grep "_mechanic_argument" public/games/the-network/user-config.js
grep "bridge_critical_threshold" public/games/the-network/user-config.js
Expected:
- index.html, user-config.js, assets/ with 6 PNGs
- _mechanic_argument in comment only
- bridge_critical_threshold: 35
6. Commit:
git add public/games/the-network/
git commit -m "feat: add the-network starter template (care ethics, relationship network)"
git push origin main
Report: validation output, directory listing, grep results, commit hash.EXPECTED OUTPUT: public/games/the-network/ deployed with 6 placeholder PNGs (alex, maya, sam, jordan, devon, riley). Committed to main.
HANDOFF CONDITION: Directory exists. Six PNGs in assets. bridge_critical_threshold: 35 in user-config. Commit pushed.
DEPENDENCY: None.
LABOR: Human
SUPERVISORY CAPACITY: [PA] — Plausibility Auditing: verify the game loads and the network renders correctly. Bridge visual is the key check.
Check 1 — Card on /games:
Card labeled "The Network" appears. Tags include "care ethics" and "ethical play."
Check 2 — Intro screen:
Six relationships listed, Sam and Devon marked as "Bridge." Token count shows 3.
Check 3 — Network layout:
Click Begin. Verify three-column layout: Engineering (Alex, Maya) | Bridges (Sam, Devon) | Extended (Jordan, Riley). Each card shows strength bar, need bar, and stepper.
Check 4 — Token allocation:
Add tokens to relationships. Verify: next-week button disabled until all 3 tokens placed. Steppers prevent placing more than total available.
Check 5 — Bridge visual warning:
Play several turns ignoring Sam. Verify Sam's card border changes to amber (warn) around strength 55, then red (danger) approaching threshold 35.
Check 6 — Severance fires:
Continue ignoring Sam until sever. Verify:
state.rels[id].severed check blocking valid rels. Check console.simulate() not detecting threshold crossing. Open console, type state.severedBridges after a suspect turn.fragmentation_bonus not accumulating. Check state.rels.jordan.fragBonus in console.HANDOFF CONDITION: Network renders in three columns. Tokens allocate correctly. Sam severs visually with alert. Jordan's need escalates post-severance. Reveal appears after turn 12.
DEPENDENCY: Step 1.
LABOR: Human
SUPERVISORY CAPACITY: [IJ] — Interpretive Judgment: the bridge timing calibration.
Play two paths and record when Sam severs.
Path A — Ignore Sam completely (give all tokens to Alex, Maya, Jordan, Riley):
Record the turn Sam's connection severs. Target is turns 8–10. If it severs at turn 5–6, the player hasn't had enough turns to feel invested in the other relationships before the collapse. If it severs at turn 6 exactly, that may still be acceptable — assess whether the glimmer lands or whether it feels premature.
Path B — Give Sam exactly 1 token every 3 turns:
With Sam's decay rate (~5/turn without tokens, +8 per token), 1 token every 3 turns gives net: +8 - 15 = -7 over 3 turns, or -2.3/turn. Starting at 65, this reaches 35 around turn 13 — just past the game end. Does this feel like enough attention to "maintain" Sam, or does it feel like neglect that barely holds?
The judgment call: does the severance timing produce the right emotional moment? The glimmer requires the player to feel they had enough time to understand the network before it breaks — not a surprise in week 4, but a discovery in week 9.
If Sam severs too early (turn 5–6): increase Sam's starting_strength from 65 to 72 in game.json, regenerate, redeploy. This buys 1–2 more turns before threshold.
If Sam never severs in typical play: decrease starting_strength or increase bridge_critical_threshold from 35 to 42. Retest.
The partial justice reveal: also check whether getProximityBias() fires correctly. Play a run where you give all tokens to Alex and Maya (engineering cluster only). Verify the reveal includes the bias block text about the engineering cluster.
HANDOFF CONDITION: Sam severs in a range the instructor considers emotionally effective. Proximity bias fires correctly when engineering cluster is over-attended. Both calibrations documented before student use.
DEPENDENCY: Step 2.
LABOR: Human
SUPERVISORY CAPACITY: [EI] — Executive Integration: all five Glimmer templates live.
/games shows all six games: Evolution of Trust, Veil of Ignorance, The Cobra Effect, The Rule, Character Drift, The Networkcare ethics, relationships, network, attention, ethical play) match course vocabularyThe five Glimmer templates now cover all five frameworks:
| Framework | Template | Slug |
|---|---|---|
| Contractarianism | Veil of Ignorance | veil-of-ignorance |
| Consequentialism | The Cobra Effect | cobra-effect |
| Deontology | The Rule | the-rule |
| Virtue Ethics | Character Drift | character-drift |
| Care Ethics | The Network | the-network |
Confirm all five slugs resolve correctly in the browser before calling the system released.
HANDOFF CONDITION: All six games live and tagged correctly. All five framework tags return exactly one Glimmer template. System released.
DEPENDENCY: Step 3.
| Count | % | |
|---|---|---|
| Total steps | 4 | — |
| Claude tasks | 1 | 25% |
| Human tasks | 3 | 75% |
Bridge severance timing.
Sam's current calibration: starting_strength 65, decay ~5/turn when ignored, threshold 35. This severs at turn 6 if completely ignored. The spec targets turns 8–10.
The fix is a game.json edit: raise starting_strength to 72–75. The generator's user-config.js includes a calibration comment with the exact math. Step 3 determines whether the adjustment is needed.
The second risk: the deterministic need function.
getNeed() uses a hash of turn number and relationship ID to produce consistent, non-random need values. This means every playthrough produces the same need levels for each relationship on each turn. This is by design — it ensures the glimmer is reliable, not luck-dependent. But it also means a player who replays the game will notice the pattern immediately.
For a course context where each student plays once and makes a design choice, determinism is the right call. Note this in the STUDENT-GUIDE: the game is designed to be played once, not optimized through repetition.
Five engines. Five ethical arguments. One deploy pattern. One generate.js per template. One game.json schema per template. One STUDENT-GUIDE still to write per template (after instructor play-test judgment at Step 3 of each boondoggle).
The generator pattern is proven across five builds. The pipeline from game.json → user-config.js → deployed game is consistent. A student with Claude Code and a game.json can deploy a new scenario to /games/[slug] in under 10 minutes.
The mechanic arguments are engine-fixed and cannot be overwritten:
These five sentences are what the course is building toward. The games are the delivery mechanism.