Skip to content

hashgraph-online/awesome-codex-plugins / plugins/alpoxdev/hypercore/skills/hyper-debate

hyper-debate

Verified

Use this skill when the user explicitly asks for adversarial multi-perspective planning before a decision-complete plan, with words such as '여러 관점으로 따져보고 계획', '찬반으로 붙여서 계획', 'adversarial planning', 'devil's advocate plan', 'red team this plan', or 'debate it then plan'. Five fixed perspectives argue through three rounds and the surviving findings become the plan. It writes per-run receipts under `.hyper/debate/`, asks for one approval of the insight bundle, then writes the plan under `.hyper/plans/`. Do not use it for an ordinary 'plan this feature' request, a review of finished code, a research report, a PRD, or an immediate edit.

FileKindSize
SKILL.ko.mdmarkdown18.9 KB
SKILL.mdskill md17.9 KB
assets/evals/hyper-debate-cases.jsonlother8.3 KB
references/plan-shape.ko.mdmarkdown9.6 KB
references/plan-shape.mdmarkdown9.3 KB
references/receipts.ko.mdmarkdown7.6 KB
references/receipts.mdmarkdown7.4 KB
rules/approval-and-plan.ko.mdmarkdown5.9 KB
rules/approval-and-plan.mdmarkdown5.8 KB
rules/perspectives.ko.mdmarkdown5.8 KB
rules/perspectives.mdmarkdown5.4 KB
rules/rounds.ko.mdmarkdown8.1 KB
rules/rounds.mdmarkdown7.8 KB

references/receipts.md

View on GitHub
# Receipts

Read this before the first dispatch and whenever a returned receipt must be validated. It owns the exact shape of every receipt, the ledger, and the approval record.

Every receipt is a JSON file written under `.hyper/debate/<slug>/`, one per perspective per round. The assignment includes the round's schema verbatim so the perspective never guesses.

## Contents

1. Common rules
2. Round 1 receipt
3. Round 2 receipt
4. Round 3 receipt
5. Ledger
6. Approval record
7. Validation checklist

## 1. Common Rules

- One file per perspective per round: `round1/<perspective>.json`, `round2/<perspective>.json`, `round3/<perspective>.json`.
- The perspective ids are exactly `skeptic`, `validator`, `researcher`, `architect`, `creative`.
- Every receipt names its `perspective` and its `round`. A receipt whose identity does not match the assignment is invalid, even when its content is good.
- Every nested object disallows additional properties. Extra keys mean the perspective invented a field; reject the receipt.
- A citation is one of: `path:line`, an official URL, `user: <quoted statement>`, or `no evidence found`. A fabricated citation invalidates the finding.
- Prose inside a receipt is Korean by default; ids, verdicts, and field names stay English.

## 2. Round 1 Receipt

```json
{
  "type": "object",
  "properties": {
    "perspective": {"type": "string", "enum": ["skeptic", "validator", "researcher", "architect", "creative"]},
    "round": {"type": "integer", "enum": [1]},
    "findings": {
      "type": "array", "minItems": 3, "maxItems": 7,
      "items": {
        "type": "object",
        "properties": {
          "id": {"type": "string"},
          "claim": {"type": "string"},
          "evidence": {"type": "array", "minItems": 1, "items": {"type": "string"}}
        },
        "required": ["id", "claim", "evidence"], "additionalProperties": false
      }
    }
  },
  "required": ["perspective", "round", "findings"], "additionalProperties": false
}
```

Check on return: 3-7 findings, ids prefixed with the perspective, at least one evidence string per finding, and no duplicate id across the run.

## 3. Round 2 Receipt

```json
{
  "type": "object",
  "properties": {
    "perspective": {"type": "string", "enum": ["skeptic", "validator", "researcher", "architect", "creative"]},
    "round": {"type": "integer", "enum": [2]},
    "critiques": {
      "type": "array", "minItems": 1,
      "items": {
        "type": "object",
        "properties": {
          "targetFindingId": {"type": "string"},
          "verdict": {"type": "string", "enum": ["ATTACK", "STANDS"]},
          "reason": {"type": "string"},
          "evidence": {"type": "array", "minItems": 1, "items": {"type": "string"}}
        },
        "required": ["targetFindingId", "verdict", "reason", "evidence"], "additionalProperties": false
      }
    }
  },
  "required": ["perspective", "round", "critiques"], "additionalProperties": false
}
```

Check on return, per receipt: exactly one judgment for every finding belonging to the other four perspectives, no unknown target id, no self-judgment, and a nonempty reason on every `STANDS`.

## 4. Round 3 Receipt

```json
{
  "type": "object",
  "properties": {
    "perspective": {"type": "string", "enum": ["skeptic", "validator", "researcher", "architect", "creative"]},
    "round": {"type": "integer", "enum": [3]},
    "responses": {
      "type": "array", "minItems": 3, "maxItems": 7,
      "items": {
        "type": "object",
        "properties": {
          "findingId": {"type": "string"},
          "verdict": {"type": "string", "enum": ["DEFEND", "REFINE", "CONCEDE"]},
          "survivingClaim": {"type": ["string", "null"]},
          "reason": {"type": "string"},
          "evidence": {"type": "array", "items": {"type": "string"}}
        },
        "required": ["findingId", "verdict", "survivingClaim", "reason", "evidence"], "additionalProperties": false
      }
    }
  },
  "required": ["perspective", "round", "responses"], "additionalProperties": false
}
```

Check on return: one response per own finding, a nonempty `survivingClaim` and nonempty `evidence` for `DEFEND` and `REFINE`, and a null `survivingClaim` for `CONCEDE`. The `reason` addresses every attack recorded against that finding.

## 5. Ledger

`.hyper/debate/<slug>/ledger.json` is a single JSON object. It is written incrementally, not once at the end.

```json
{
  "request": {"verbatim": "", "optIn": "", "scope": [], "outcome": ""},
  "slug": "",
  "mode": "full | bundle-only | review",
  "execution": "parallel | sequential",
  "reduced": false,
  "reducedReason": null,
  "grounding": [{"path": "", "note": ""}],
  "waves": [
    {"round": 1, "perspectives": [{"id": "skeptic", "job": "", "attempt": 1, "receipt": "round1/skeptic.json"}]}
  ],
  "findings": [
    {
      "id": "skeptic-1",
      "perspective": "skeptic",
      "claim": "",
      "evidence": [""],
      "critiques": [{"by": "validator", "verdict": "ATTACK", "reason": ""}],
      "response": {"verdict": "REFINE", "survivingClaim": "", "reason": ""},
      "decision": "keep | refine | open-question | drop",
      "decisionReason": ""
    }
  ],
  "surviving": {"skeptic": 0, "validator": 0, "researcher": 0, "architect": 0, "creative": 0},
  "dropped": {"skeptic": 0, "validator": 0, "researcher": 0, "architect": 0, "creative": 0},
  "openQuestions": [{"question": "", "why": "", "gates": []}],
  "approval": {"requestedAt": "", "request": "", "decidedAt": null, "decision": null},
  "terminal": "planned | awaiting-approval | reduced-planned | declined | blocked",
  "notes": []
}
```

## 6. Approval Record

The approval request and its answer live in the ledger under `approval`, and the presented text is saved at `.hyper/debate/<slug>/approval.md` so the reader can see exactly what was approved.

```markdown
# <slug> - Approval

## Decisions presented
- <decision>

## Risks and mitigations
- <risk> -> <mitigation>

## Open questions
- <question> -> <who must decide>

## Asked
<the exact question put to the user>

## Decision
<the user's answer, verbatim, or `no answer`>
```

An empty or missing `## Decision` body means no approval.

## 7. Validation Checklist

Before advancing past any round:

- [ ] Five receipts exist for the round, one per perspective.
- [ ] Each receipt's `perspective` and `round` match its filename and the assignment.
- [ ] No receipt carries a key outside its schema.
- [ ] Round 1: 3-7 findings each, every finding cited, ids unique across the run.
- [ ] Round 2: full coverage of other-perspective findings, no self-judgment, no unknown id, a reason on every `STANDS`.
- [ ] Round 3: one response per own finding, evidence on every `DEFEND`/`REFINE`, a null claim on every `CONCEDE`, every attack addressed.
- [ ] The ledger records the wave identities, attempts, and receipt paths.

## Sources

> Checked 2026-11-10.

| Claim | Source |
|---|---|
| The round-1, round-2, and round-3 receipt schemas | `dbc-hbin/omp-ulw@74df40f hyperplan/references/rounds.md` (mechanism reference, re-authored) |
| Coverage over schema as the round-2 check, and the own-finding check in round 3 | `dbc-hbin/omp-ulw@74df40f hyperplan/references/rounds.md` (mechanism reference, re-authored) |
| The ledger fields, the approval record, and the validation checklist | this package's chosen design, 2026-11-10 (issue #26) |

### Evidence grade

`LOCAL` - the schemas are re-authored from a repository with no license; no text or code is copied. The ledger, approval record, and checklist are this package's design.
reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis.Stars: 469kUpdated Claude CodeCodexVerifiedInstall
diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness.Stars: 322kUpdated Claude CodeCodexVerifiedInstall
dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependenciesStars: 322kUpdated Claude CodeCodexVerifiedInstall
executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is availableStars: 322kUpdated Claude CodeCodexVerifiedInstall