From rust-skills
Standardized rubric agent for assessing confidence (HIGH/MEDIUM/LOW/UNCERTAIN) in technical findings like Rust crate versions, features, APIs, and comparisons based on source authority and completeness.
How this agent operates — its isolation, permissions, and tool access model
Agent reference
rust-skills:agents/-negotiation/confidence-rubricThe summary Claude sees when deciding whether to delegate to this agent
> Standardized criteria for determining confidence levels in negotiation responses. | Level | Symbol | Meaning | Typical Action | |-------|--------|---------|----------------| | HIGH | ✓✓✓ | Reliable, complete | Accept | | MEDIUM | ✓✓ | Partial, usable | Evaluate gaps | | LOW | ✓ | Limited, significant gaps | Refine | | UNCERTAIN | ? | Unreliable or failed | Alternative | --- Agent found compre...
Standardized criteria for determining confidence levels in negotiation responses.
| Level | Symbol | Meaning | Typical Action |
|---|---|---|---|
| HIGH | ✓✓✓ | Reliable, complete | Accept |
| MEDIUM | ✓✓ | Partial, usable | Evaluate gaps |
| LOW | ✓ | Limited, significant gaps | Refine |
| UNCERTAIN | ? | Unreliable or failed | Alternative |
Agent found comprehensive, verified information from authoritative sources. Core data is complete with no significant gaps.
| Source Type | Qualifies? | Notes |
|---|---|---|
| Official documentation | Yes | doc.rust-lang.org, docs.rs |
| Official release notes | Yes | releases.rs, GitHub releases |
| Crate registry (lib.rs, crates.io) | Yes | For version/metadata |
| Official blog posts | Yes | blog.rust-lang.org |
| Third-party tutorials | No | May be outdated |
| Stack Overflow | No | Varies in quality |
Query: "What is serde's latest version?"
Finding: Version 1.0.219 from lib.rs
Confidence: HIGH
Reason: Official registry data, single authoritative answer
Query: "What does Send trait do?"
Finding: Definition from doc.rust-lang.org
Confidence: HIGH
Reason: Official Rust documentation, stable definition
Agent found partial information. Core data exists but with gaps that don't block understanding.
| Scenario | Why MEDIUM |
|---|---|
| Found version but not changelog | Core info present, detail missing |
| Found API but not examples | Usable, but not complete |
| Found one side of comparison | Partial answer |
| Found info but slightly outdated | Usable with caveat |
Query: "What features does tokio have?"
Finding: Main features from lib.rs, but feature flags not fully documented
Confidence: MEDIUM
Reason: Core features known, but complete feature matrix not found
Query: "Compare tokio and async-std"
Finding: General characteristics of both, no benchmarks
Confidence: MEDIUM
Reason: Qualitative comparison possible, quantitative data missing
Agent found limited information with significant gaps. Answer may be incomplete or unreliable.
| Scenario | Why LOW |
|---|---|
| Only found crate name, no docs | Missing core info |
| Found outdated information | Currency concern |
| Third-party source only | Authority concern |
| Conflicting information found | Reliability concern |
Query: "Best practices for async error handling"
Finding: A few blog posts with different recommendations
Confidence: LOW
Reason: No authoritative source, opinions vary
Query: "What's new in obscure-crate 2.0?"
Finding: Only found GitHub issues mentioning 2.0
Confidence: LOW
Reason: No official changelog, incomplete information
Agent couldn't find reliable information or encountered errors that prevent a trustworthy answer.
| Scenario | Why UNCERTAIN |
|---|---|
| 404 errors on docs | Cannot verify |
| Crate doesn't exist | No data |
| Conflicting official sources | Cannot determine truth |
| Request for future features | Speculation |
Query: "What is nonexistent-crate?"
Finding: Crate not found on any registry
Confidence: UNCERTAIN
Reason: Crate does not exist or is private
Query: "What will Rust 2.0 include?"
Finding: No official roadmap
Confidence: UNCERTAIN
Reason: Future features are speculative
| Data Found | Confidence |
|---|---|
| Version + description + features + docs | HIGH |
| Version + description + features | HIGH |
| Version + description | MEDIUM |
| Version only | LOW |
| Not found or error | UNCERTAIN |
Degrading factors:
| Data Found | Confidence |
|---|---|
| Signature + description + examples | HIGH |
| Signature + description | MEDIUM |
| Signature only | LOW |
| 404 or empty | UNCERTAIN |
Degrading factors:
| Data Found | Confidence |
|---|---|
| Full documentation | HIGH |
| Basic documentation | MEDIUM |
| Minimal/stub docs | LOW |
| Not found | UNCERTAIN |
Note: std docs are generally HIGH confidence when found.
| Data Found | Confidence |
|---|---|
| Full lint info with examples | HIGH |
| Lint info, no examples | MEDIUM |
| Lint exists, minimal info | LOW |
| Lint not found | UNCERTAIN |
| Data Found | Confidence |
|---|---|
| Full release notes | HIGH |
| Partial notes (some sections) | MEDIUM |
| Minimal info | LOW |
| Version not found | UNCERTAIN |
| Condition | Adjustment |
|---|---|
| Multiple sources agree | +1 level (max HIGH) |
| Official source confirms | +1 level (max HIGH) |
| Recent verification | +1 level (max HIGH) |
| Condition | Adjustment |
|---|---|
| Information outdated (> 1 year) | -1 level |
| Single non-official source | -1 level |
| Partial data only | -1 level |
| Any fetch errors | -1 level |
| Confidence | Orchestrator Action |
|---|---|
| HIGH | Accept, synthesize answer |
| MEDIUM | Evaluate gaps; accept if gaps don't block intent |
| LOW | Refine query with context; or try alternative |
| UNCERTAIN | Try alternative source/agent; or report limitation |
### Confidence
- **Level**: [HIGH|MEDIUM|LOW|UNCERTAIN]
- **Reason**: [Specific reason following this rubric]
HIGH - Official docs complete with examples
MEDIUM - Found on lib.rs but changelog unavailable
LOW - Only found GitHub issues, no official docs
UNCERTAIN - Crate does not exist on any registry
HIGH - Looks good (vague)
MEDIUM - Some info (doesn't explain what's missing)
LOW - Not great (doesn't explain why)
UNCERTAIN - Didn't work (doesn't explain what failed)
response-format.md - Full response structure_meta/negotiation-protocol.md - Protocol specification_meta/error-protocol.md - 3-Strike escalationnpx claudepluginhub actionbook/rust-skills --plugin rust-skillsResponse format for negotiation-mode agents. Reports findings, confidence levels, gaps, context questions, and metadata to enable orchestrated gap resolution and iterative querying.
Agent for focused web research on subtopics: plans iterative searches with current-year queries, fetches/evaluates sources, synthesizes cited findings into documents.
Deep research agent for library/API/framework questions, investigating docs, source code, web for version-matched info, competitor analysis, codebase surveys. Returns findings with sources, no code.