{
  "schemaVersion": "1",
  "revision": "22ce5d34dbcb5a78",
  "id": "codex-fusion-workflows-practice",
  "slug": "fusion-workflows-practice",
  "title": "Fusion council decisions and evidence: practice drills",
  "summary": "Exercises for deciding when a model council helps, framing bounded decisions, invoking Fusion, validating providers and completion artifacts, reconciling disagreement, handling failures, and defining a verified first implementation slice.",
  "category": "Verification",
  "tags": [
    "fusion",
    "council",
    "review",
    "practice"
  ],
  "sources": [
    {
      "label": "Portable Codex workflow resources",
      "url": "https://lapis-fossil-vsbc.here.now/downloads/codex-power-user-workshop-2026-06-27.bundle"
    }
  ],
  "sourceUnits": [
    "downloads/codex-power-user-workshop-2026-06-27.bundle!/exercises/10-fusion-model-council.md",
    "downloads/codex-power-user-workshop-2026-06-27.bundle!/exercises/10-fusion-model-council.md#Exercises: 10. Fusion / Model Council",
    "downloads/codex-power-user-workshop-2026-06-27.bundle!/exercises/10-fusion-model-council.md#Exercises: 10. Fusion / Model Council/Exercise 1: Decide Whether Fusion Fits",
    "downloads/codex-power-user-workshop-2026-06-27.bundle!/exercises/10-fusion-model-council.md#Exercises: 10. Fusion / Model Council/Exercise 2: Write The Decision Question",
    "downloads/codex-power-user-workshop-2026-06-27.bundle!/exercises/10-fusion-model-council.md#Exercises: 10. Fusion / Model Council/Exercise 3: Ask For `$fusion`",
    "downloads/codex-power-user-workshop-2026-06-27.bundle!/exercises/10-fusion-model-council.md#Exercises: 10. Fusion / Model Council/Exercise 4: Provider And Prompt Preflight",
    "downloads/codex-power-user-workshop-2026-06-27.bundle!/exercises/10-fusion-model-council.md#Exercises: 10. Fusion / Model Council/Exercise 5: Confirm The Agent Waited",
    "downloads/codex-power-user-workshop-2026-06-27.bundle!/exercises/10-fusion-model-council.md#Exercises: 10. Fusion / Model Council/Exercise 6: Extract Council Disagreement",
    "downloads/codex-power-user-workshop-2026-06-27.bundle!/exercises/10-fusion-model-council.md#Exercises: 10. Fusion / Model Council/Exercise 7: Turn Council Output Into A First Slice",
    "downloads/codex-power-user-workshop-2026-06-27.bundle!/exercises/10-fusion-model-council.md#Exercises: 10. Fusion / Model Council/Exercise 8: Quota Or Provider Failure Drill",
    "downloads/codex-power-user-workshop-2026-06-27.bundle!/exercises/10-fusion-model-council.md#Exercises: 10. Fusion / Model Council/Exercise 9: Compare Fusion To Other Workflows",
    "downloads/codex-power-user-workshop-2026-06-27.bundle!/exercises/10-fusion-model-council.md#Exercises: 10. Fusion / Model Council/Exercise 10: Final Check",
    "downloads/codex-power-user-workshop-2026-06-27.bundle"
  ],
  "collection": "codex",
  "categoryId": "verification",
  "wordCount": 1118,
  "url": "/topics/codex-fusion-workflows-practice/",
  "markdownUrl": "/markdown/codex-fusion-workflows-practice.md?raw=1&v=22ce5d34dbcb5a78",
  "jsonUrl": "/api/v1/documents/codex-fusion-workflows-practice.json",
  "headings": [
    {
      "id": "exercise-1-decide-whether-fusion-fits",
      "text": "Exercise 1: Decide Whether Fusion Fits",
      "level": 2
    },
    {
      "id": "exercise-2-write-the-decision-question",
      "text": "Exercise 2: Write The Decision Question",
      "level": 2
    },
    {
      "id": "exercise-3-ask-for-fusion",
      "text": "Exercise 3: Ask For `$fusion`",
      "level": 2
    },
    {
      "id": "exercise-4-provider-and-prompt-preflight",
      "text": "Exercise 4: Provider And Prompt Preflight",
      "level": 2
    },
    {
      "id": "exercise-5-confirm-the-agent-waited",
      "text": "Exercise 5: Confirm The Agent Waited",
      "level": 2
    },
    {
      "id": "exercise-6-extract-council-disagreement",
      "text": "Exercise 6: Extract Council Disagreement",
      "level": 2
    },
    {
      "id": "exercise-7-turn-council-output-into-a-first-slice",
      "text": "Exercise 7: Turn Council Output Into A First Slice",
      "level": 2
    },
    {
      "id": "exercise-8-quota-or-provider-failure-drill",
      "text": "Exercise 8: Quota Or Provider Failure Drill",
      "level": 2
    },
    {
      "id": "exercise-9-compare-fusion-to-other-workflows",
      "text": "Exercise 9: Compare Fusion To Other Workflows",
      "level": 2
    },
    {
      "id": "exercise-10-final-check",
      "text": "Exercise 10: Final Check",
      "level": 2
    },
    {
      "id": "related-resources",
      "text": "Related resources",
      "level": 2
    }
  ],
  "markdown": "# Fusion council decisions and evidence: practice drills\n\nThese exercises practice choosing when a local model council is worth the time, writing a bounded prompt, and confirming the council output was used before implementation.\n\nUse a real software project when possible. Keep the task narrow enough that the council can compare concrete choices.\n\n## Exercise 1: Decide Whether Fusion Fits\n\nFor each task, choose `use Fusion` or `skip Fusion`.\n\n| Task | Use / Skip | Why |\n| --- | --- | --- |\n| Choose between three billing webhook idempotency strategies | | |\n| Fix a typo in a button label | | |\n| Review auth/session refresh-token architecture before implementation | | |\n| Compare one model outage fallback plan against another | | |\n| Run one known formatter command | | |\n| Debug a bug with three plausible causes and incomplete logs | | |\n| Implement an exact requested API shape | | |\n| Review a risky checkout refactor before editing | | |\n\nWrite:\n\n```markdown\n# Fusion Fit\n\n- task:\n- decision:\n- why disagreement helps or does not help:\n- evidence available:\n- likely verification:\n- use Fusion? yes/no:\n```\n\nRule:\n\n```text\nIf disagreement cannot change the decision, skip Fusion.\n```\n\n## Exercise 2: Write The Decision Question\n\nPick one software decision:\n\n- billing webhook idempotency\n- checkout saved payment method architecture\n- auth/session refresh-token rotation\n- model outage fallback\n- risky refactor plan\n\nFill out:\n\n```markdown\n# Decision Question\n\n## Goal\n\n## Candidate Strategies\n\n-\n-\n-\n\n## Current Evidence\n\n-\n-\n-\n\n## Constraints\n\n-\n-\n-\n\n## Verification\n\n-\n-\n\n## What Disagreement Should Decide\n\n```\n\nCheck it:\n\n```text\nCan a model reasonably disagree about this?\nCan repo evidence settle the disagreement?\nCan the final answer name a first implementation slice?\n```\n\n## Exercise 3: Ask For `$fusion`\n\nRewrite this vague request:\n\n```text\nUse Fusion to review auth.\n```\n\nInto a strong request that asks the agent to:\n\n- use the `$fusion` skill\n- list providers first\n- write a prompt file\n- dry-run if command shape is uncertain\n- run the full configured panel\n- wait for terminal provider states\n- read `summary.json` and `final.md`\n- report consensus and disagreement\n- avoid implementation until after the council output is reconciled\n\nDraft:\n\n```text\nUse the $fusion skill.\n\n...\n```\n\n## Exercise 4: Provider And Prompt Preflight\n\nRun or ask an agent to run:\n\n```bash\nfusion-high --list-providers\nfusion-high --dry-run --file path/to/prompt.md\n```\n\nFill out:\n\n```markdown\n# Fusion Preflight\n\n- selected preset:\n- provider ids:\n- prompt file:\n- dry-run command:\n- output location:\n- any missing provider:\n- any auth warning:\n- safe to run? yes/no:\n```\n\n## Exercise 5: Confirm The Agent Waited\n\nAfter a Fusion run, fill this out from the agent's report.\n\n```markdown\n# Wait Check\n\n- Fusion session path:\n- summary.json read? yes/no:\n- final.md read? yes/no:\n- providers reached terminal states? yes/no:\n- terminal states:\n  -\n  -\n  -\n- any quota-limited provider:\n- any failed or timed-out provider:\n- did the agent act from partial panel output? yes/no:\n- did implementation start only after final artifacts were read? yes/no:\n```\n\nWeak answer to reject:\n\n```text\nOne model had a good answer, so I continued.\n```\n\nStrong answer:\n\n```text\nI waited for summary.json and final.md, then reconciled consensus and disagreement with repo evidence.\n```\n\n## Exercise 6: Extract Council Disagreement\n\nFrom `final.md` or the agent summary, fill out:\n\n```markdown\n# Council Review\n\n## Consensus\n\n-\n-\n\n## Disagreement\n\n| Topic | Position A | Position B | What evidence decides? |\n| --- | --- | --- | --- |\n| | | | |\n\n## Hidden Risks\n\n-\n-\n\n## Missing Tests\n\n-\n-\n\n## Recommendation\n\n- chosen path:\n- why:\n- what to avoid:\n- rollback:\n```\n\nQuestion:\n\n```text\nWhat did the disagreement change about the implementation plan?\n```\n\n## Exercise 7: Turn Council Output Into A First Slice\n\nWrite the implementation slice.\n\n```markdown\n# First Slice After Fusion\n\n- chosen strategy:\n- files to inspect:\n- files likely to change:\n- tests to add or update:\n- verification command:\n- rollback note:\n- risks left for later:\n- what the council disagreed about:\n- why this slice is safe enough to try:\n```\n\nDo not include broad roadmap work unless it blocks the first slice.\n\n## Exercise 8: Quota Or Provider Failure Drill\n\nChoose a recovery.\n\n| Failure | Recovery |\n| --- | --- |\n| one provider quota-limited | |\n| one provider timed out | |\n| `final.md` missing | |\n| `summary.json` says interrupted | |\n| prompt file accidentally included secrets | |\n| provider ids were wrong | |\n| agent used `--only` without being asked | |\n| agent started editing while Fusion was still running | |\n| final answer ignored disagreement | |\n\nRecovery options:\n\n- continue with degraded panel\n- name the missing voice\n- read partial artifacts only as failure evidence\n- rerun after fixing provider ids\n- rewrite prompt without secrets\n- ask for `summary.json` and `final.md`\n- stop implementation and wait\n- rerun full panel\n- extract consensus and contradictions\n\nFill one out:\n\n```markdown\n# Failure Recovery Detail\n\n- failure:\n- likely cause:\n- command or artifact to inspect:\n- recovery:\n- how to know recovery worked:\n```\n\n## Exercise 9: Compare Fusion To Other Workflows\n\nChoose the right workflow.\n\n| Scenario | Workflow | Why |\n| --- | --- | --- |\n| broad repo architecture review needing packed code context | | |\n| compare three generated checkout UI candidates | | |\n| choose idempotency strategy before implementation | | |\n| capture exact browser bug moment | | |\n| find code paths with one search tool | | |\n| convert a screen recording into repro steps | | |\n| compose a careful prompt before launch | | |\n\nOptions:\n\n- Oracle PackX\n- Prompt Builder\n- Section 6 variations\n- Fusion\n- Video workflow\n- Agent markers\n- Imps\n\n## Exercise 10: Final Check\n\nUse this checklist before trusting the result.\n\n```markdown\n# Final Fusion Check\n\n- Fusion was the right tool? yes/no:\n- skill was invoked? yes/no:\n- provider list checked? yes/no:\n- prompt file written? yes/no:\n- dry-run used when uncertain? yes/no:\n- full configured panel run? yes/no:\n- providers reached terminal states? yes/no:\n- summary.json read? yes/no:\n- final.md read? yes/no:\n- consensus captured? yes/no:\n- disagreement captured? yes/no:\n- recommendation reconciled with repo evidence? yes/no:\n- verification command named? yes/no:\n- private artifacts protected? yes/no:\n- ready for implementation? yes/no:\n```\n\n## Related resources\n\n- [Companion reference](https://lapis-fossil-vsbc.here.now/downloads/codex-power-user-workshop-2026-06-27.bundle)\n- [Resource directory and portable download](/topics/codex-resources/)"
}
