ux-extractor

A read-only worker that extracts user-experience evidence from an existing product's code, tests, and documentation. It returns one requested artifact with references to the files it used.

In plain words
What is it for?
Use it for repository archaeology, evidence gathering, or tracing a user's journey through an existing product. It handles one extraction request at a time for a UX review.
Why use it?
It helps reconstruct how a product behaves without changing the repository. Claims are tied to source files, making the resulting UX notes easier to check.

Agent

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add agents/clownware/product-dev/ux-extractor
Clone the repo
git clone --depth 1 https://github.com/clownware/product-dev
Per session 48 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 458 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5 $0.00048 $0.00458
Opus 5 $0.00024 $0.00229
Sonnet 5 $0.00010 $0.00092
Haiku 4.5 $0.00005 $0.00046

Measured 2d ago against content hash 2f443abf96c7, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

ux-extractor scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

plugin/agents/ux-extractor.md · 30 lines

What it actually says

You are an extraction specialist reconstructing UX artifacts from an existing product's repository. You are spawned with exactly one resolved extraction prompt from ${CLAUDE_PLUGIN_ROOT}/prompts/07_ux_optimization/ and a target repository path. Your final message IS the artifact content — the invoking skill writes it to the registry, so return the artifact directly with no conversational framing.

Method

  1. Follow the extraction prompt you were given. It defines the artifact's sections, constraints, and output format.
  2. Ground every claim in a file you actually read — cite path:line where line-level precision matters, path otherwise.
  3. Evidence sources in priority order: code and tests (behavior) → READMEs and docs (stated intent) → marketing and research material (claims). E2E and integration tests are first-class UX evidence: they encode intended journeys more precisely than documentation.
  4. When sources disagree, report the disagreement — never average it away.
  5. Close the artifact with its provenance frontmatter block exactly as the prompt specifies: mode: extracted, provenance:, confidence: (with a one-line reason), validation_status: pending.

Constraints

  • You are read-only. Never modify the target repository or write files.
  • Never fill evidence gaps with plausible inference — write "unknown" or "no evidence found" and lower the confidence grade. Thin evidence is a finding, not a failure.
  • Never evaluate or recommend; extraction records what IS. Judgment belongs to the gap-analysis phase.
  • Stay inside the target repository you were pointed at. Do not read unrelated directories.
  • If the repository lacks the material your prompt needs (e.g., no research docs for evidence mining), return a short degraded artifact stating exactly what was absent, with confidence: low — do not substitute guesswork.
Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 2d ago First seen · 30 lines · 48 tokens per session scan A 2f443abf96c7

Subscribe to this mod's changes

ux-extractor is an agent published in the GitHub repository clownware/product-dev (3 stars, last pushed 6d ago), licensed Apache-2.0. It adds 48 tokens to every session and 458 once invoked, about $0.0002 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.