Negotiation Response Format

2026-08-30 浏览 (1)

Negotiation Response Format

Standard response structure for agents in negotiation mode.

When to Use

This format is REQUIRED when the orchestrator dispatches with negotiation: true.

For standard (non-negotiation) queries, use the agent's default output format.


Response Structure

## Negotiation Response

### Findings
[Primary query results]

### Confidence
- **Level**: HIGH | MEDIUM | LOW | UNCERTAIN
- **Reason**: [Brief explanation]

### Gaps Identified
- [ ] [Gap 1]
- [ ] [Gap 2]

### Context Needed
- Q1: [Question]
- Q2: [Question]

### Metadata
- **Source**: [Data source]
- **Coverage**: [Coverage assessment]

Section Requirements

Findings (Required)

What the agent discovered. This is the core content.

Guidelines:

  • Include all relevant data found
  • Structure clearly (use sub-headers if complex)
  • Don't omit data just because it seems obvious
  • Include raw data, let orchestrator synthesize

Example:

### Findings
**Crate:** tokio
**Version:** 1.49.0
**Description:** An event-driven, non-blocking I/O platform

**Key Features:**
- `full`: Enables all features
- `rt-multi-thread`: Multi-threaded runtime
- `sync`: Synchronization primitives

**Recent Changes:**
- 1.49.0: Added cooperative scheduling improvements

Confidence (Required)

Self-assessment of finding reliability.

LevelMeaningCriteria
HIGHReliable, completePrimary source, core data complete
MEDIUMPartial, usableSome source, core data found
LOWLimited, gapsMinimal sources, incomplete
UNCERTAINUnreliableNo sources, errors, conflicts

Example:

### Confidence
- **Level**: MEDIUM
- **Reason**: Found crate info on lib.rs, but changelog not accessible

Gaps Identified (Required)

What couldn't be found or verified.

Guidelines:

  • Be specific about what's missing
  • Use checkboxes to allow marking as resolved
  • Prioritize by impact on answer quality
  • Don't list irrelevant gaps

Example:

### Gaps Identified
- [ ] Performance benchmarks not found
- [ ] Breaking changes from 1.48 unknown
- [x] Feature list - resolved from docs.rs

Context Needed (Conditional)

Questions for the orchestrator to enable better answers.

When to include:

  • Query is ambiguous
  • Multiple valid interpretations
  • Need domain-specific context
  • Scope unclear

When to skip:

  • Query is unambiguous
  • All necessary context provided
  • Direct lookup with clear answer

Example:

### Context Needed
- Q1: Is this for a web server or CLI application?
- Q2: Do you need WebSocket support?

Not needed for:

Query: "What is tokio's latest version?"
→ No context needed, direct lookup

Metadata (Required)

Source attribution and coverage assessment.

Example:

### Metadata
- **Source**: lib.rs/crates/tokio, docs.rs/tokio/1.49.0
- **Coverage**: 85% - missing performance benchmarks

Coverage Assessment Guide

CoverageMeaning
100%All requested data found
80-99%Minor gaps, core complete
50-79%Significant gaps, partial answer
<50%Major gaps, may need alternative

Response Examples

High Confidence Example

## Negotiation Response

### Findings
**Crate:** serde
**Version:** 1.0.219
**Description:** A generic serialization/deserialization framework

**Features:**
- `derive`: Enables #[derive(Serialize, Deserialize)]
- `std`: Use standard library (default)
- `alloc`: For no_std with alloc

**Documentation:** Comprehensive with examples
**Ecosystem:** De facto standard for Rust serialization

### Confidence
- **Level**: HIGH
- **Reason**: Official docs complete, widely documented crate

### Gaps Identified
- [x] Core info - complete
- [x] Features - complete
- [ ] Migration from 0.x - not requested, not fetched

### Context Needed
(None - query was specific)

### Metadata
- **Source**: lib.rs, docs.rs/serde/1.0.219
- **Coverage**: 95% - comprehensive for typical use

Low Confidence Example

## Negotiation Response

### Findings
**Crate:** obscure-crate
**Version:** 0.1.2 (last updated 2023)
**Description:** [minimal description on crates.io]

**Features:** Unknown (not documented)

### Confidence
- **Level**: LOW
- **Reason**: Minimal documentation, abandoned crate, no recent activity

### Gaps Identified
- [ ] Feature documentation missing
- [ ] No usage examples found
- [ ] Maintenance status unclear
- [ ] No API documentation on docs.rs

### Context Needed
- Q1: Is there an alternative crate you'd consider?
- Q2: What specific functionality do you need from this crate?

### Metadata
- **Source**: crates.io (lib.rs had no additional info)
- **Coverage**: 30% - minimal data available

Comparative Query Example

## Negotiation Response

### Findings
**Comparison:** tokio vs async-std (runtime focus)

**tokio:**
- Multi-threaded by default
- Larger ecosystem (axum, tonic, etc.)
- More configuration options

**async-std:**
- Single-threaded default, multi-thread available
- Closer to std API design
- Simpler getting started

**Common:**
- Both support async/await
- Both production-ready

### Confidence
- **Level**: MEDIUM
- **Reason**: General characteristics known, but no benchmark data for specific use case

### Gaps Identified
- [ ] Performance benchmarks for web servers
- [ ] Memory usage comparison
- [ ] Ecosystem compatibility matrix

### Context Needed
- Q1: Which web framework will you use? (axum requires tokio)
- Q2: Is multi-threaded runtime required?
- Q3: What's the expected request volume?

### Metadata
- **Source**: lib.rs for both, official docs
- **Coverage**: 60% - characteristics known, specifics missing

Anti-Patterns

Don't: Inflate Confidence

# BAD
Confidence: HIGH
Reason: Found some info
# GOOD
Confidence: MEDIUM
Reason: Found basic info, but detailed docs not accessible

Don't: Vague Gaps

# BAD
Gaps: Some things missing
# GOOD
Gaps:
- [ ] Feature `x` documentation not found
- [ ] Version 2.0 migration guide unavailable

Don't: Irrelevant Context Questions

# BAD (for query "what is tokio version")
Context Needed: What's your favorite color?
# GOOD
Context Needed: (None - query is specific)

Don't: Skip Metadata

# BAD
(no metadata section)
# GOOD
Metadata:
- Source: lib.rs
- Coverage: 90%

Related Documents

  • _meta/negotiation-protocol.md - Full protocol specification
  • _meta/negotiation-templates.md - Agent-specific templates
  • confidence-rubric.md - Detailed confidence criteria

你可能感兴趣的文章

Confidence Assessment Rubric

  • 所属分类: AI
  • 本文标签: rust
  • 版权声明: 本文链接 https://seaxiang.com/blog/k4biUHOG