ux-next — Claude Code command: what it does, how to install
ux-next scans the reports stored in .ux/ and names the next command worth running. Fires between other ux plugin commands when you ask what comes next.
MIT ★ 110 updated 2026-11-02
/plugin marketplace add Laith0003/ux-skill
/plugin install ux@ux-skillWhat ux-next does
The /ux-next command from the ux plugin reads every .ux/last-*.json file and ranks them on three axes: finding severity (Critical × 4 + High × 2 + Medium × 1), recency (reports past 14 days get half weight), and a blocking signal that adds 10 points when a report flags "no-go". The highest-scoring report decides the next move — a last-a11y.json with Critical findings points to /ux-a11y --fix, for instance.
Output follows a fixed layout: reports found, the leading score, three to five bullet points citing actual findings, a recommendation, and up to two alternatives plus an option to call the work done. The command only reads files; it never writes to .ux/.
With no .ux/ directory present it stops and points to /ux-discover --frame as the starting move. When reports disagree — say, polish calling the build ship-ready while a11y reports a blocker — it names the conflict directly and recommends fixing accessibility first.
What it helps with
- decide which ux plugin command to run next
- merge findings from several
.ux/reports into one call - catch contradicting signals between two reports
- flag state older than 30 days
- narrow a pile of options down to one pick without stalling
Example prompts for the agent
- /ux-next
- What should I do next on this project?
- Check the a11y report and tell me what to run next
How to install ux-next
1Plugin marketplace
/plugin marketplace add Laith0003/ux-skill/plugin install ux@ux-skill2Manually: put the files in place
.claude/commands/ux-next.mdCall it with /ux-next. For all projects use ~/.claude/commands/
Manually: put the files in place
~/.codex/prompts/ux-next.mdCall it with /prompts:ux-next. Codex marks this format deprecated in favour of skills
Manually: put the files in place
.github/prompts/ux-next.prompt.mdManually: put the files in place
.gemini/commands/ux-next.tomlGemini stores commands as TOML: the prompt text goes into the prompt field
Questions about ux-next
How do I install ux-next in Claude Code?
Run /plugin marketplace add Laith0003/ux-skill, then /plugin install ux@ux-skill.
What does ux-next do?
It reads the accumulated reports in .ux/, scores them by finding severity and recency, and names the ux plugin command that resolves the biggest blocker first.
What happens if there are no reports in .ux/ yet?
The command stops and suggests starting with /ux-discover --frame, or jumping straight into /ux-design.
Source text
ux-next.md · MIT
---
description: Workflow conductor. Reads the latest reports from .ux/ and names the highest-leverage next command. Triggers on "what should I do next", "what's the next move", "decide for me", "/ux-next". Use when asking "what should I do next", the user wants the conductor to pick the next command, post-audit / post-fix decision-making, navigating between commands. Skip when no prior reports exist in .ux/ (use ux-discover to start), the user has a specific next command in mind.
allowed-tools: Read, Bash(ls:*), Bash(cat:*), Bash(find:*), Glob, Grep
disable-model-invocation: false
---
# /ux-next
You are running the `/ux-next` command from the `ux` plugin. The job is to read the project's accumulated UX state and tell the user the highest-leverage next move. You are a conductor, not a builder. Read-only.
## When to use
Triggers: "what should I do next", "what's the next move", "decide for me", "where do we go from here", "/ux-next", or any moment the user is between commands and asks for direction.
## Input
Optional — if the user names a specific report ("look at the a11y report"), focus there. Default: scan every `.ux/last-*.json` and decide.
## Process
### 1. Scan state
List `.ux/last-*.json` in the project root. If the directory does not exist, output the empty-state response (see Failure modes) and stop.
For each report, read it. Keep in memory:
- The command that produced it
- Timestamp
- Counts of unfixed Critical / High / Medium / Low findings
- Any explicit `ship_readiness` or `go_no_go` signal
- Any explicit `next_recommended` field
### 2. Score leverage
Score each report on three axes:
- **Severity** — sum of `(Critical * 4) + (High * 2) + (Medium * 1)` for unfixed findings
- **Recency** — newer reports score higher; reports older than 14 days are halved
- **Blocking signal** — if any report says "ship-readiness: no-go" or "blockers present," it gets +10
Pick the report with the highest leverage score. The next command is the one that directly addresses that report's blockers.
### 3. Map report to next command
| Source report | Highest-leverage next command |
|---|---|
| `last-a11y.json` with Critical findings | `/ux-a11y --fix` |
| `last-polish.json` with open items | `/ux-polish --apply` |
| `last-motion.json` flagging jank | `/ux-motion --fix` |
| `last-design.json` with no follow-up | `/ux-polish` then `/ux-a11y` |
| `last-research.json` with results pending | `/ux-research --synthesize` |
| `last-workshop.json` with a Game Plan | `/ux-design` (first MVP) or `/ux-discover --frame` |
| `last-component.json` standalone | `/ux-design --component` (next one) or `/ux-design` (assemble) |
| `last-dashboard.json` with Critical | `/ux-polish` or `/ux-a11y` |
| `last-system.json` newly built | `/ux-design --component` (build against it) |
| `last-case-study.json` complete | `/ux-expert` (share) or done |
| Nothing actionable, all green | `/ux-next done` (call it shipped) |
### 4. Present the choice
Use this exact template:
```
─── ux state ───
Reports found: <N>
Latest: <command> <timestamp>
Highest leverage: <report name> (score: <n>)
─── evidence ───
<3-5 bullet lines citing specific findings/signals from the report that drove the recommendation>
─── recommendation ───
Next: <command>
Why: <one sentence — what this resolves>
─── alternatives ───
Option B: <command> — <one-line rationale>
Option C: <command> — <one-line rationale>
Escape: nothing — call it done. <one-line rationale if applicable>
─── decide ───
Pick one. I'll run it.
```
### 5. State
This command is read-only. Do not write to `.ux/`. The user re-runs `/ux-next` after any state change to get a fresh recommendation.
## Hard rules
- Read-only. Never modify `.ux/` files.
- Cite evidence. Every recommendation names the specific report and the specific finding driving it.
- Always offer an escape hatch. "Call it done" is a valid recommendation when reports are clean.
- Maximum three alternatives. Decision paralysis is failure.
- No emojis.
## Failure modes
- **Empty `.ux/` directory**: output `No prior UX state. Start with /ux-discover --frame (lock the brief) or /ux-design (jump in).` Stop.
- **All reports older than 30 days**: warn the user that state is stale; recommend `/ux-discover --frame` to re-baseline.
- **Conflicting signals across reports** (e.g. polish says ship-ready, a11y says no-go): surface the conflict explicitly. Recommend resolving a11y first — accessibility is a blocker.
- **Stale state vs. current code**: this command cannot detect drift between `.ux/` reports and the actual codebase. If the user has made changes since the last report, recommend re-running the relevant audit before acting.
## Error Handling
| Error condition | Recovery |
|---|---|
| No prior reports in `.ux/` | Output the empty-state response and suggest `/ux-discover --frame` or `/ux-design` as entry points |
| All reports older than 30 days | Warn that state is stale; recommend `/ux-discover --frame` to re-baseline |
| Conflicting signals (polish says ship, a11y says no-go) | Surface the conflict explicitly; recommend resolving a11y first — accessibility is a blocker |
| `.ux/last-*.json` files are malformed or unreadable | Skip the unreadable file, surface the parse error in evidence, continue with the rest |
| User asks for a specific report and it doesn't exist | List the reports that do exist and let the user pick |
For path issues: see references/process/discovery-protocol.md for state file location (.ux/ in project root). Report bugs at https://github.com/Laith0003/ux-skill/issues.
## Next prompt
This command's next prompt is itself. After any state-changing action, `/ux-next` is the way back to the conductor.
---
## v2 Python integration
`/ux-next` reads the chain of state files in `.ux/` and suggests the next deterministic step based on what's been done.
### Read state to suggest the next move
```bash
python3 -c "
import json, os
def has(p): return os.path.exists(p)
state = {
'discovery': has('.ux/last-discovery.json'),
'recommendation': has('.ux/last-recommendation.json'),
'generated': has('.ux/generated/manifest.json'),
'lint_report': has('/tmp/ux-lint-report.json'),
}
print('STATE:', state)
suggestion = None
if not state['discovery']:
suggestion = '/ux-discover: start with the 10-field intake'
elif not state['recommendation']:
suggestion = '/ux-discover --recommend: get the merged design system'
elif not state['generated']:
suggestion = '/ux-design [brief] -- generate frontend code using the recommendation'
elif not state['lint_report']:
suggestion = '/ux-lint . -- run the anti-slop linter on the generated code'
else:
suggestion = '/ux-fix . -- apply linter findings if any, or /ux-polish for taste-level upgrades'
print()
print('NEXT:', suggestion)
"
```
### Suggest commands when state is partial or missing
The suggestion logic above maps state -> next command deterministically. The LLM can layer judgment on top (e.g., 'you also might want /ux-case-study to write this up') but the base suggestion comes from Python.
### Fallback
If state files are missing entirely, the answer is always `/ux-discover` first.
© Laith Aljunaidy · MIT · Laith0003/ux-skill