Atlas / Skills / nwave-ai / Nw Document

Nw DocumentSAFE

skills/nwave-ai/nw-document

AI agents that guide you from idea to working code, with you in control at every step.

Verdict
SAFE
Grade
B
Trust score
89 /100
Version
—
Hosts
—
License
MIT
Stars
617
01

Overview

AI agents that guide you from idea to working code, with you in control at every step.

Read from source at commit 1a864ec81cdeOBSERVED · 2026-10-14
Advertisement
02

What it tells the agent

The instruction file, verbatim from the audited commit — this is the text the model reads, and the surface the audit's instruction layer examines. Quoted here so you can judge it without cloning anything.

---
name: nw-document
description: "Creates evidence-based documentation following DIVIO/Diataxis principles. Use when writing tutorials, how-to guides, reference docs, or explanations."
user-invocable: true
argument-hint: '[topic/component] - Optional: --type=[tutorial|howto|reference|explanation] --research-depth=[overview|detailed|comprehensive|deep-dive]'
---

# NW-DOCUMENT: DIVIO Documentation Creation

**Wave**: CROSS_WAVE | **Agent**: Orchestrator (self)
**Agents**: Nova (nw-researcher)|Scholar (nw-researcher-reviewer)|Quill (nw-documentarist)|nw-documentarist-reviewer

## Overview

Create evidence-based, DIVIO-compliant documentation by orchestrating research and writing phases with peer review at each gate. Cross-wave capability for any nWave phase.

## Context Files Required

- .nwave/trusted-source-domains.yaml — Embed inline in researcher prompt
- ~/.claude/agents/nw/nw-researcher.md — Extract research methodology
- ~/.claude/agents/nw/nw-documentarist.md — Extract DIVIO framework and templates

## Command Syntax

```bash
/nw-document [topic] [--type=tutorial|howto|reference|explanation] [--research-depth=overview|detailed|comprehensive|deep-dive]
```

If `--type` omitted, ask user. If `--research-depth` omitted, auto-select: tutorial->overview|howto->detailed|reference->comprehensive|explanation->deep-dive.

## Orchestration Phases

Sub-agents have no Skill tool access. Embed all domain knowledge inline in each Task prompt. Read from agent files and config at orchestration time.

```
Phase 1: Research           @nw-researcher
Phase 1.5: Research Review  @nw-researcher-reviewer
Phase 2: Documentation      @nw-documentarist
Phase 2.5: Doc Review       @nw-documentarist-reviewer
Phase 3: Handoff
```

### Phase 0: Pre-Flight

1. Validate topic non-empty, type/depth valid if provided
2. If type not specified, present DIVIO selection: TUTORIAL ("Teach me")|HOW-TO ("Help me do X")|REFERENCE ("What is X?")|EXPLANATION ("Why is X?")
3. Determine output: `docs/{tutorials|howto|reference|explanation}/{topic-kebab-case}.md`
4. Read and cache: .nwave/trusted-source-domains.yaml|nw-researcher.md|nw-documentarist.md

### Phase 1: Research (@nw-researcher)

Invoke via Task tool. Prompt includes: topic|doc type|research depth|complete .nwave/trusted-source-domains.yaml (inline)|type-specific research focus (from nw-researcher.md)|quality gates: trusted sources only, 3+ sources/claim, citation coverage >95%, source reputation >=0.80. Output: `docs/research/{topic-kebab-case}-for-{type}-doc.md`

### Phase 1.5: Research Review (@nw-researcher-reviewer)

Prompt includes: research artifact path|review focus (source verification|bias detection|evidence quality|cross-reference|doc readiness)|quality gates (same as Phase 1)|output: append YAML review metadata. Verdicts: APPROVED|NEEDS_REVISION|REJECTED

### Phase 2: Documentation (@nw-documentarist)

Prompt includes: topic|type|research path|DIVIO framework principles (from nw-documentarist.md)|type-specific validation rules + template|collapse detection anti-patterns|quality gates: type purity >=80%, Flesch 70-80, zero spelling errors, zero broken links, style >=95%. Output: doc file + .validation.yaml

### Phase 2.5: Documentation Review (@nw-documentarist-reviewer)

Prompt includes: doc artifact path|review focus (classification accuracy|validation completeness|collapse detection|recommendation quality|scores|verdict)|quality gates (same as Phase 2)|output: append YAML review metadata. Verdicts: APPROVED|NEEDS_REVISION|RESTRUCTURE_REQUIRED

### Phase 3: Handoff

Verify all deliverables exist and both reviews APPROVED. Present: research path|doc path|validation path|quality gate results|review outcomes and iteration count.

## Review Iteration Protocol

Applies to both review gates (Phase 1.5 and 2.5):
1. **APPROVED**: Proceed to next phase
2. **NEEDS_REVISION**: Extract critiques, re-invoke producer with feedback, re-invoke reviewer. Max 2 cycles. Escalate to user after 2 failures.
3. **REJECTED / RESTRUCTURE_REQUIRED**: Escalate immediately — restart|adjust scope|accept with issues|cancel.

## Examples

### Example 1: Tutorial with auto-depth
```bash
/nw-document "Getting Started with nWave" --type=tutorial
```
Orchestrator auto-selects overview depth, invokes researcher, reviews, invokes documentarist, reviews, outputs to `docs/tutorials/getting-started-with-nwave.md`.

### Example 2: Explanation with explicit depth
```bash
/nw-document "Why nWave Uses Hexagonal Architecture" --type=explanation --research-depth=deep-dive
```
Full 4-phase pipeline with deep-dive research. Output to `docs/explanation/`.

### Example 3: Interactive type selection
```bash
/nw-document "Mikado Method Integration"
```
Orchestrator prompts user to select DIVIO type before proceeding.

## Success Criteria

- [ ] Research created with trusted sources and cross-references
- [ ] Research review: APPROVED (max 2 cycles)
- [ ] Documentation follows DIVIO template with >=80% type purity
- [ ] Doc review: APPROVED (max 2 cycles)
- [ ] All deliverables exist: research|documentation|validation report
- [ ] No collapse anti-patterns detected

## Error Handling

- Insufficient sources: continue with gaps|expand scope|cancel
- Review iteration limit exceeded: escalate with persistent issues
- Collapse detected: split into separate docs|revise scope|accept

## Next Wave

**Handoff To**: Invoking workflow
**Deliverables**:
- Research: `docs/research/{topic}-for-{type}-doc.md`
- Documentation: `docs/{type-dir}/{topic-kebab-case}.md`
- Validation: `docs/{type-dir}/{topic-kebab-case}.md.validation.yaml`
03

Trust audit

SAFEgrade B · trust 89/100 Nothing in the source contradicts what it says it does. Grade A is reserved for packages that have also passed the behavioural sandbox.

LayerWhat it checksResult
L0Provenance & inventoryPASS
L1Static analysis of the codeNA
L2Instruction surface (what it tells the agent)PASS
L3Class-specific surfacePASS
L4Behavioural (sandbox)SKIPPED

What the source does

Filesystem
none-observed
Network
none-observed
Shell
none-observed
Dependencies
pinned
Secrets in source
none-found

Findings (0)

No findings outside the package's declared scope.

Gates applied: no_behavioural_pass.

Audited 2026-10-14 · audit v0.4.1 · source sha 1a864ec81cdefull audit observations/trust-audit/skill/nwave-ai__nw-document.json · Report an issue / request a re-scan
04

Audit history

Every audit this skill has had.

DateSourceVerdictGradeScoreChange
2026-10-141a864ec81cdeSAFEB89first audit
05

Questions

What does the Nw Document skill do?

AI agents that guide you from idea to working code, with you in control at every step.

Is Nw Document safe to install?

The audit found nothing in the source that contradicts what it says it does, and graded it B (89/100). Grade A is held back for packages that have also passed a sandboxed behavioural run, which is why a clean skill reads B.

What can Nw Document access on my machine?

The audit observed no filesystem, network or shell use at all in its source.

How current is this page?

The grade is for one exact copy of the source (1a864ec81cde), read on 2026-10-14. The repository is watched, and a new audit runs when it changes — this is the first audit.

Advertisement