Atlas / Skills / nwave-ai / Nw Roadmap

Nw RoadmapSAFE

skills/nwave-ai/nw-roadmap

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-roadmap
description: "Creates a phased roadmap.json for a feature goal with acceptance criteria and TDD steps. Use when planning implementation steps before execution."
user-invocable: false
argument-hint: '[agent] [goal-description] - Example: @solution-architect "Migrate to microservices"'
---

# NW-ROADMAP: Goal Planning

**Wave**: CROSS_WAVE
**Agent**: Architect (nw-solution-architect) or domain-appropriate agent

## Overview

Dispatches expert agent to fill a pre-scaffolded YAML roadmap skeleton. CLI tools handle structure; agent handles content.

Output: `docs/feature/{feature-id}/deliver/roadmap.json`

## Usage

```bash
/nw-roadmap @nw-solution-architect "Migrate monolith to microservices"
/nw-roadmap @nw-software-crafter "Replace legacy authentication system"
/nw-roadmap @nw-product-owner "Implement multi-tenant support"
```

## Execution Steps

You MUST execute these steps in order. Do NOT skip any.

1. **Parse Parameters** — Extract agent name (after @, validated against agent registry), goal description (quoted string), and derive feature-id from goal in kebab-case (e.g., "Migrate to OAuth2" -> "migrate-to-oauth2"). Gate: agent name, goal, and feature-id all resolved.

2. **Scaffold Skeleton** — Run `des.cli.roadmap init` via Bash BEFORE invoking agent. Gate: CLI exits 0; stop and report error on non-zero exit.

```bash
des-roadmap init \
  --project-id {feature-id} \
  --goal "{goal-description}" \
  --output docs/feature/{feature-id}/deliver/roadmap.json
```
For complex projects add: `--phases 3 --steps "01:3,02:2,03:1"`

Do NOT write the file manually.

3. **Invoke Agent** — Invoke the named agent via Task tool to fill skeleton TODO placeholders. Gate: agent completes without error.

```
@{agent-name}

Fill in the roadmap skeleton at docs/feature/{feature-id}/deliver/roadmap.json.
Replace every TODO with real content. Do NOT change the YAML structure
(phases, steps, keys). Fill in: names, descriptions, acceptance criteria,
time estimates, dependencies, and implementation_scope paths.

Goal: {goal-description}

# Test paradigm mandate (STANDING 2026-05-05) — apply per test level

GUARD — emit by deliverable_type (ADR-PST-003 / OPEN-5-A): the mandate
block below applies ONLY when `deliverable_type == application` (or when
deliverable_type is unset — application is the default). For a plugin/skill
deliverable, do NOT emit the mandate; instead include this one-line directive
in each step's `implementation_notes`:

  STRUCTURAL VERIFICATION (mandatory, plugin/skill): This step ships prompt
  text / config / docs, not application code, so the Hypothesis + state-delta
  unit mandate does not apply. Verify the deliverable structurally — a golden
  assertion (skill-lint) over the changed markdown/config asserting the
  required block is present and correctly shaped — gated at CI + pre-push.

When `deliverable_type == application` (or unset), for EVERY step's
`implementation_notes`, include this directive verbatim unless the step is
documented as exempt:

  TEST PARADIGM (mandatory): Unit tests for this step MUST be written as
  property-based tests using Hypothesis @given strategies + state-delta
  matchers (assert_state_delta over a declared universe with strict=True).
  Single-example unit tests are FALLBACK only — document why in a
  `# bypass:` comment if used. Acceptance scenarios MUST be framed as
  Property: blocks (quantified preconditions + invariant outcomes) where
  the contract supports it; classic Scenario: is fallback. Integration
  and E2E tests REMAIN single-example — they verify wiring, not
  invariants. Universe declaration is mandatory even when only one slot
  is checked: the unchanged() predicate covers the rest.

  STAGE CONTEXT (cascade): NEW code = Stage 0 (paradigm-from-day-zero,
  no debt accumulates). Bug-prone legacy code = Stage 1 first (state-
  delta migration captures hidden mutations, 33-75% yield), THEN Stage 2
  PBT amplification (low yield by design — surface is debt-paid). Don't
  expect high hit rate on Stage-1-completed surfaces; that's healthy.
  See nw-tdd-methodology::Paradigm Mandate for the full debt-payoff
  framing.

When a step's contract genuinely cannot be expressed as a property
(e.g., golden-file diff, exact error string, single-shot UI flow),
include in `implementation_notes`:

  EXEMPT FROM PARADIGM: <one-sentence reason>. Compensate with
  property-based test at <unit|integration> level on adjacent slot.
```

Context to pass (if available): measurement baseline|mikado-graph.md|existing docs.

4. **Validate** — Run `des-roadmap validate` via Bash. Gate: exit 0 = success; exit 1 = print errors and stop; exit 2 = usage error, stop.

```bash
des-roadmap validate docs/feature/{feature-id}/deliver/roadmap.json
```

## Invocation Principles

Keep agent prompt minimal. Agent knows roadmap structure and planning methodology.

Pass: skeleton file path + goal description + measurement context (if available).
Do not pass: YAML templates|phase guidance|step decomposition rules.

For performance roadmaps, include measurement context inline so agent can validate targets against baselines.

## Success Criteria

### Dispatcher (you) — all 4 must be checked

- [ ] 1. Parameters parsed (agent name, goal, feature-id)
- [ ] 2. `des.cli.roadmap init` executed via Bash (exit 0)
- [ ] 3. Agent invoked via Task tool to fill TODO placeholders
- [ ] 4. `des.cli.roadmap validate` executed via Bash (exit 0)

### Agent output (reference)

- [ ] 5. All TODO placeholders replaced with real content
- [ ] 6. Steps are self-contained and atomic
- [ ] 7. Acceptance criteria are behavioral and measurable
- [ ] 8. Step decomposition ratio <= 2.5 (steps / production files)
- [ ] 9. Dependencies mapped, time estimates provided

## Error Handling

- Invalid agent: report valid agents and stop
- Missing goal: show usage syntax and stop
- Scaffold failure (exit 2): report CLI error and stop
- Validation failure (exit 1): print errors, do not proceed

## Examp
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-roadmap.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 Roadmap skill do?

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

Is Nw Roadmap 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 Roadmap 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