# Analyze

> Find conflicts between spec, plan, tasks and code at any point (SDD diagnostic)

- Skill: `janszewczyk/analyze` (Agent Skill)
- Install (CLI): `npx get-skill add janszewczyk/analyze`
- Raw SKILL.md: https://api.skillmd.com/api/skills/janszewczyk/analyze/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: JanSzewczyk (https://skillmd.com/u/janszewczyk)
- Updated: 2026-11-15
- Page: https://skillmd.com/skills/janszewczyk/analyze

---


# SDD Analyze

A read-only diagnostic: where do the spec, plan, tasks and code disagree **right now**? Use it
mid-feature, after changing the spec, or before `/sdd:review`. It edits nothing and runs no tests.

## Steps

### 1. Verify

Resolve the feature: the `specs/*/spec.md` whose `**Branch:**` equals `git branch --show-current`
(none → list the specs and ask). Invoke `spec-verifier` once (Agent tool) with the feature folder and
the **feature diff** as defined in [feature-diff.md](${CLAUDE_PLUGIN_ROOT}/references/feature-diff.md) (generated globs excluded).

### 2. Classify

Combine its result with the task statuses in `tasks.md`:

- an unmet AC whose task is `draft` or `in-progress` → **progress** (not done yet, not an error);
- an unmet AC whose task is `review` or `done`, or that no task serves → **gap**;
- out-of-scope changes, non-goal violations and every `drift` item → **findings**.

### 3. Report

Short, in chat: AC met / pending / gaps, then the findings, at most about 10, each with `file:line`.
For each, the next move:

- unmet AC of an implemented task → `/sdd:implement <task>`
- plan or tasks do not cover an AC → edit `plan.md`, then `/sdd:tasks`
- ambiguous or outdated spec → `/sdd:clarify`
- nothing wrong → `/sdd:review` when every task is `review|done`

Quote task ids, AC ids and file paths; never flag the generated globs.

