# Evaluation Criteria & Process

### AI Oracle Prompting Standards

Seedify uses multiple AI oracles (Claude, ChatGPT, Gemini, Grok) with standardized prompts.

Standard Prompt Structure:

```
You are evaluating a blockchain project milestone for completion.

PROJECT: [Name]
MILESTONE: [Title]
DEADLINE: [Date]

COMPLETION CRITERIA:
[List criteria with targets]

EVIDENCE SUBMITTED:
[Evidence package with links]

EVALUATION TASKS:
1. Verify each criterion is met using provided evidence
2. Check for evidence manipulation (wash trading, Sybil attacks, fake metrics)
3. Assess evidence quality and credibility
4. Flag any ambiguous or borderline cases
5. Provide binary recommendation: PASS or FAIL

OUTPUT FORMAT:
{
  "recommendation": "PASS" | "FAIL",
  "criteria_results": [
    {
      "criterion": "[text]",
      "met": true | false,
      "confidence": "high" | "medium" | "low",
      "evidence_quality": "strong" | "adequate" | "weak",
      "notes": "[explanation]"
    }
  ],
  "red_flags": ["list of concerns"],
  "overall_assessment": "[2-3 sentences]"
}
```

### Seedify Manual Review

After AI Oracle evaluation, Seedify team conducts manual review.

Review Checklist:

* [ ] All evidence links are functional and verifiable
* [ ] On-chain data matches claimed metrics
* [ ] No obvious wash trading or Sybil activity
* [ ] Evidence timestamps are within milestone period
* [ ] Quality meets project standards (not just "technically complete")
* [ ] AI oracles did not hallucinate or misinterpret data
* [ ] No conflicts with previous milestone decisions (consistency)

Review Outcomes:

1. Agree with Oracle (Pass): Recommend approval to Community 
2. Agree with Oracle (Fail): Recommend denial to Community 
3. Disagree with Oracle (False Positive): Override with disclaimer, recommend denial
4. Disagree with Oracle (False Negative): Override with disclaimer, recommend approval

Seedify Disclaimer Format (when overriding Oracle):

```
⚠️ SEEDIFY OVERRIDE NOTICE

The AI Oracle evaluation resulted in [PASS/FAIL].

Seedify has reviewed the evidence and determined this is a [FALSE POSITIVE / FALSE NEGATIVE].

REASONING:
[Detailed explanation of why Oracle was incorrect]

SEEDIFY RECOMMENDATION: [APPROVE / DENY]

The Community may choose to follow Oracle or Seedify recommendation.
```

### Community Vote Presentation

All Community votes receive a standardized report.

Report Structure:

```markdown
# Milestone [X] Evaluation Report
**Project**: [Name]
**Deadline**: [Date]
**Submission**: [On Time / Late by X hours]

## Oracle Results
- Claude: [PASS/FAIL]
- ChatGPT: [PASS/FAIL]
- Gemini: [PASS/FAIL]
- Grok: [PASS/FAIL]

**Oracle Consensus**: [4/4 Pass | 3/4 Pass | Split | Fail]

## Seedify Review
**Recommendation**: [APPROVE / DENY / ABSTAIN]
**Review Confidence**: [High / Medium / Low]

## Completion Criteria Analysis
[Table showing each criterion, evidence, and pass/fail]

## Evidence Quality Assessment
- **On-Chain Verification**: [% of criteria verified on-chain]
- **Evidence Credibility**: [High / Medium / Low]
- **Red Flags**: [None / List]

## Key Findings
[3-5 bullet points of most important observations]

## Voter Guidance
- **Complexity**: [Low / Medium / High] - How technical is this milestone?
- **Ambiguity**: [Low / Medium / High] - How clear-cut is pass/fail?
- **Risk Level**: [Low / Medium / High] - Confidence in evaluation

## Full Evidence Package
[Link to complete submission]
```


---

# Agent Instructions: Querying This Documentation

If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.

Perform an HTTP GET request on the current page URL with the `ask` query parameter:

```
GET https://docs.seedify.fund/milestone-guidance-criteria/evaluation-criteria-and-process.md?ask=<question>
```

The question should be specific, self-contained, and written in natural language.
The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.