convez.ai
Skill
speech-works

sw-guardrails

speech-works/sw-guardrails

Ver no GitHub
BomPluginScore · de 0 a 100

A Claude Code skill that reviews Speechworks content and code for therapeutic integrity and brand voice in stuttering and stammering work. Under active development.

Instalação

$npx convez-ai speech-works/sw-guardrails

1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.

Métricas

Estrelas
0
Forks
0
Acompanhando
0
Saúde
75%
Último commit
há 1 mês
Versão
—
Issues abertas
0
Contribuidores
1

README

A tradução para o português deste README ainda está na fila.

sw-guardrails

Status: under active development. The guidelines, severities, and grep patterns will keep changing as we learn. Tagged releases for version pinning are on the roadmap; until then, pin a specific commit if you need stable behaviour.

A Claude Code skill, installable as a plugin or used as a standalone checklist, for reviewing any Speechworks artifact against the therapeutic-integrity and brand guidelines that govern how we write about stuttering and stammering.

It checks words and code alike: blog posts, marketing and landing copy, UI strings, AI Coach and other LLM prompts, analytics event names, metric and KPI names, dashboard tiles, schema fields, and the copy buried inside source files.

[!IMPORTANT] Found something that looks wrong? This skill is about a sensitive, lived experience, so it will contain mistakes, blind spots, and framing that does not fit everyone. Please help us correct it. See CONTRIBUTING.md for a copy-paste issue format and a short process. If a guideline contradicts your lived experience, you are right and the rule needs to change. Corrections from people who stutter or stammer, SLPs, and families are especially welcome.


Why this exists

Speechworks builds practice tools for people who stutter, and for the speech-language pathologists who work beside them. The risk in that work is easy to miss: a product, or a single sentence, that treats fluency as the goal can quietly harm the people it is meant to help.

This is a long-standing consensus among the clinicians who built the field, most of whom stuttered themselves:

  • Stuttering works like an iceberg. The blocks and word-swaps a listener notices are the tip; the larger mass below the surface is shame, fear, guilt, avoidance, isolation, and the wish to be cured. That mass is what keeps people stuck. (Sheehan, ch. 6; Molt, ch. 21.)
  • Trying not to stutter usually makes it worse. The way out runs through it: accepting it, avoiding less, telling people, and learning to stutter more easily rather than less. (Ramig, ch. 9; Moses, ch. 15.)
  • There is no cure. The goal is a changed relationship with stuttering. (Van Riper, ch. 28; Neely, ch. 12.)
  • Success means approach, not fluency: situations entered, disclosures made, avoidance dropped, risks taken. David Daly counted his attempts by moving paperclips from one pocket to the other. (Manning, ch. 22; Daly, ch. 16.)

So the rule everything else hangs on, the "fluency god" guardrail, is simple: never imply a cure, and never make fluency the headline metric. Let fluency be a side effect, and measure approach instead.

A fuller write-up of the audience mindset, and how it maps to product and content decisions, lives in the Speechworks team notes. This repo is the enforceable version.


What it checks

The full ruleset lives in SKILL.md. In summary:

SectionFocusExample catch
A. Therapeutic integritycure claims, fluency-as-metric, avoidance or suppression framing, shame, relapse, honest tone"Become 100% fluent"; a fluencyScore KPI; "take a deep breath and the words will come"
B. Voice and framingpeer authority, lived detail, the Coach as a companion rather than a judge, the dual PWS/SLP audiencea faceless-AI "we'll fix you" tone
C. Languageboth stuttering and stammering; person-first in brand voice; no pity or othering"sufferer", "afflicted", "victim"
D. Codemetric, event, and variable names; dashboards; LLM prompts; embedded copy; schema enumsevent stutter_detected; toast "Great, no stutters detected!"

Each finding comes with a severity (🔴 block, 🟡 warn, 🔵 note), a reason, and a concrete rewrite.


Install and use

Using a different AI tool (Cursor, a custom agent, an LLM app)? See Use with any AI agent → for every install method and the raw skill URL.

As a Claude Code plugin (recommended)

This repo is also a plugin marketplace. Add it once, then install the plugin:

/plugin marketplace add speech-works/sw-guardrails
/plugin install sw-guardrails@speechworks

To auto-enable it for everyone working in a given repo, commit this to that repo's .claude/settings.json:

{
  "extraKnownMarketplaces": {
    "speechworks": {
      "source": { "source": "github", "repo": "speech-works/sw-guardrails" }
    }
  },
  "enabledPlugins": ["sw-guardrails@speechworks"]
}

When a teammate clones that repo and trusts it, Claude Code registers the marketplace and enables the plugin automatically. Updates come with /plugin marketplace update speechworks.

As a project skill (no plugin)

Or drop it straight into a project's skills directory:

# from your project root
git clone https://github.com/speech-works/sw-guardrails.git \
  .claude/skills/sw-guardrails

That gives you .claude/skills/sw-guardrails/SKILL.md, which Claude Code picks up automatically. Update it later with git -C .claude/skills/sw-guardrails pull, or pin to a fixed commit instead of tracking main:

git -C .claude/skills/sw-guardrails checkout <commit-sha>

Either way

In a session:

/sw-guardrails review web/src/app/page.tsx
/sw-guardrails is this landing copy on-message?
/sw-guardrails audit our analytics event names

You can also read SKILL.md as a plain checklist for PR reviews or editorial sign-off, with no Claude Code involved. The grep quick-reference at the bottom makes a fast first pass for a CI step.


Status and roadmap

  • Core guidelines (A to D), severities, and the grep quick-reference
  • Packaged as a Claude Code plugin and marketplace
  • Submit to the Claude Community Marketplace so it is discoverable to all Claude Code users
  • Worked before-and-after examples for each rule
  • A CI-friendly grep script (scripts/check.sh)
  • Test fixtures (known-good and known-bad samples) to catch drift
  • Better detection of person-first wording versus the reclaimed word "stutterer"
  • Tagged releases and a changelog once the rules settle

Issues and pull requests are welcome, especially from people who stutter and from SLPs. If a rule here ever contradicts lived experience, the lived experience wins, so please open an issue.


References

These guidelines blend several independent authorities (US, UK, and Indian), so they don't rest on one source or one era. They draw on the sources below and quote them briefly, with attribution. Quotations are short and used for commentary; all rights stay with the original publishers.

[!NOTE] We did not adopt everything we found in these sources. Some ideas were deliberately left out (a few are even debated within the communities that hold them). For full transparency, What we did not adopt lists them in a table with sources and our reasoning.

  1. Hood, S. B. (Ed.). (2008). Advice to Those Who Stutter (2nd ed., Publication No. 0009). Stuttering Foundation of America. ISBN 0-933388-39-X. Twenty-eight first-person chapters by speech-language pathologists who stutter themselves. PDF: https://www.stutteringhelp.org/sites/default/files/Migrate/book0009_may2010.pdf. Chapters drawn on directly:
    • Ch. 1: Lon L. Emerick, Express Yourself or Go by Freight (admit the need to change; drop the "excess baggage").
    • Ch. 2: Dorvan H. Breitenfeldt, Managing Your Stuttering… (advertising, or acknowledging that you stutter; eye contact).
    • Ch. 3: Margaret Rainey, Stuttering: What You Can Do About It (you are your own worst critic).
    • Ch. 4: Hugo H. Gregory, Two Sides of the Coin (ERA-SM; the stutter-more-fluently approach).
    • Ch. 6: Joseph G. Sheehan, Message to a Stutterer (the iceberg; "you can stutter your way out of this problem").
    • Ch. 7: Frederick P. Murray, Toward Freer Speech (you have a choice about how you stutter).
    • Ch. 8: James L. Aten, Overcoming Fear and Tension (positive planning instead of anticipated failure).
    • Ch. 9: Peter R. Ramig, Don't Ever Give Up! (the most disabling part comes from our attempts not to stutter).
    • Ch. 10: J. David Williams, Basic Goals for a Person Who Stutters.
    • Ch. 12: Margaret M. Neely, Suggestions for Self-Therapy (nothing cures it; you can manage it).
    • Ch. 13: Henry Freund, Self-Improvement After Unsuccessful Treatments (relapse is expected; a trusted companion helps).
    • Ch. 14: Harold L. Luper, Some Helpful Attitudes… (realistic goals; establish, transfer, then maintain).
    • Ch. 15: Gerald R. Moses, Message to Adult Stutterers (reduce avoidances; judge yourself by whether you approached the situation).
    • Ch. 16: David A. Daly, …Gaining and Sustaining Improved Fluency (count attempts; the honest clinician who promised nothing).
    • Ch. 17: Joseph G. Agnello, Change: Potential Qualities Become Actualities (fake stuttering and the diner story; moving from reactive to proactive).
    • Ch. 18: Richard M. Boehmler, Four Steps to Freedom (identify, plan, work, maintain).
    • Ch. 19: Bill Murphy, Recovery Journal (shame stories; normalizing).
    • Ch. 20: Sol Adler, Face Your Fears.
    • Ch. 21: Larry Molt, Attacking the Iceberg… (forgiveness, understanding, courage, patience; the "pimple" analogy, where the flaw looms larger to you than to anyone else).
    • Ch. 22: Walter H. Manning, Finding Your Own Path Without Professional Help (the "things I do because I stutter" list; "don't chase the fluency god").
    • Ch. 23: Paul E. Czuchna, Guidelines (the covert versus overt stutterer).
    • Ch. 24: Robert W. Quesal, Knowledge, Understanding, and Acceptance.
    • Ch. 25: Gary J. Rentschler, Maintaining Dignity While Living With Stuttering (explaining your stuttering to a listener; the wheelchair analogy).
    • Ch. 26: Kenneth O. St. Louis, Your Life Is Too Important to Spend It Worrying About Stuttering.
    • Ch. 27: Harold B. Starbuck, Do-It-Yourself Kit for Stutterers ("the kit was empty: you already have the tools").
    • Ch. 28: Charles Van Riper, Putting It Together ("there are no magical cures… you can change your abnormal reactions").
  2. Stuttering Foundation of America. Myths and facts, brochures, and clinician referral lists. https://www.stutteringhelp.org
  3. The Stuttering Homepage (ed. Judy Kuster), a long-running directory of stuttering resources, listed in the source book at stutteringhomepage.com.
  4. National Stuttering Association (NSA). https://www.westutter.org. No cure; "frequency of stuttering is not a meaningful measure of progress"; disclosure that creates understanding rather than apologises; anti-shame; and concrete therapist red flags. Pages used include understanding-stuttering, myths/causes, find-speech-therapist, and a-new-approach-to-self-disclosure.
  5. The Indian Stammering Association (TISA). https://stammer.in. The India lens: acceptance-first ("हकलाओ मगर प्यार से", stammer with love), peer self-help over clinical dependency, and a clear warning against predatory "cure" advertising. Used for the India-context guidance.
  6. STAMMA, the British Stammering Association. https://stamma.org. Uses "stammering"; source of the RCSLT-endorsed editorial/language guidelines (words to use and to avoid), the framing "it's not the stammer we're trying to fix, it's the struggle," and "apps and devices won't cure stammering" and are "helpful for some people but not others." Pages: editorial-guidelines, finding-right-words, is-there-a-cure, apps-fluency-devices.
  7. ASHA (American Speech-Language-Hearing Association). https://www.asha.org/practice-portal/clinical-topics/fluency-disorders/. The clinical anchor: "increasing fluency may not be a goal," "stuttering itself does not inherently require remediation," no valid cures, and a note that many prefer neuroaffirming language.
  8. Stammering-pride / neurodiversity scholarship. The "Did I Stutter?" project (https://didistutter.org), Christopher Constantino's "stuttering gain" research, and Stammering Pride and Prejudice: Difference not Defect (eds. Campbell, Constantino, Eldridge, 2019), read alongside its critiques (a strong genetic/neurological basis; the risk of pressuring people in either direction). Grounds the balanced self-determination rule.

The rule about carrying both spellings, "stuttering" and "stammering", is a Speechworks decision about search and audience (UK and India search behaviour). The difference/condition framing and the words-to-avoid list draw on STAMMA and ASHA; the India context draws on TISA. See the Speechworks SEO notes.


Disclaimer

This skill captures communication and brand guidance distilled from the sources above. It is not clinical advice, and it is not a substitute for assessment or therapy by a qualified speech-language pathologist. Any misreading of the cited work is the maintainers' alone.

License

MIT © Speechworks.

Ver o original no GitHub

Bom para

Relacionados