Skip to content
Back to search

attention-control

MarketplaceWorks with Claude CodeGitHub

Air traffic control discipline for agent output, written for a reader with ADHD: action first, one meaning per word, state restated every turn.

Safety scan

A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.

Checked today
Badge for your README
findsafeskills safety rating

Markdown

[![findsafeskills safety rating](/api/listings/aaddrick-attention-control/badge.svg)](/listings/aaddrick-attention-control)

HTML

<a href="/listings/aaddrick-attention-control"><img src="/api/listings/aaddrick-attention-control/badge.svg" alt="findsafeskills safety rating" /></a>

Install

Install instructions from the README

As written by the repo's author. findsafeskills has not verified these steps.

<details open> <summary><strong>Claude Code</strong></summary>

Claude Code is the only agent here with a native output style slot. Run these two commands in your terminal:

claude plugin marketplace add aaddrick/attention-control
claude plugin install attention-control@attention-control

Then run `/config`, select **Output style**, and pick **Attention Control**. It takes effect after `/clear` or the next session.

To skip the picker, add `outputStyle` to `~/.claude/settings.json`. It is a top-level key. It does not go inside `env`, `permissions`, or any other block:

{
  "model": "opus",
  "env": { "EXAMPLE_VAR": "1" },
  "outputStyle": "Attention Control"
}

`model` and `env` stand in for keys you may already have. Keep them. Add the `outputStyle` line beside them.

For one session instead of every session, use the skill the plugin also ships:

/attention-control:attention-control

Say "stop attention control" to turn it off.

</details>

<details> <summary><strong>Codex</strong></summary>

Codex has no output style slot, so the rules ship as a skill.

codex plugin marketplace add aaddrick/attention-control --ref main
codex plugin add attention-control@attention-control

Inside Codex, `/plugins` opens the plugin browser instead.

Start a new thread, then type the skill:

$attention-control:attention-control

Codex prefixes a plugin skill with the plugin name. Say "stop attention control" to turn it off. For every turn instead, put the always-on snippet from [INSTALL.md](./INSTALL.md#the-always-on-snippet) in `~/.codex/AGENTS.md`.

</details>

<details> <summary><strong>Cursor, Gemini CLI, Copilot, Zed, and manual install</strong></summary>

See [INSTALL.md](./INSTALL.md). None of them has an output style slot, so the rules ship as a skill, a rules file, or an `AGENTS.md` block. On a skill route, type `/attention-control` and say "stop attention control" to turn it off.

</details>

What was scanned

The files the safety scan read, as written by the repository's author and shown as plain text. Lines with a finding are marked.

.claude-plugin/marketplace.json · Manifest · 24 lines

View this file on GitHub (read at the repository's current version: the scan did not record which commit it read)

  1. {
  2. "$schema": "https://www.schemastore.org/claude-code-marketplace.json",
  3. "name": "attention-control",
  4. "description": "Air traffic control discipline for agent output, written for a reader with ADHD: action first, one meaning per word, state restated every turn.",
  5. "owner": {
  6. "name": "aaddrick",
  7. "url": "https://github.com/aaddrick"
  8. },
  9. "plugins": [
  10. {
  11. "name": "attention-control",
  12. "description": "Output style that shapes every response for a reader with ADHD and writes each sentence in controlled English.",
  13. "source": "./",
  14. "category": "productivity",
  15. "keywords": [
  16. "adhd",
  17. "output-style",
  18. "accessibility",
  19. "simplified-technical-english",
  20. "productivity"
  21. ]
  22. }
  23. ]
  24. }
README.md · README · 293 lines

View this file on GitHub (read at the repository's current version: the scan did not record which commit it read)

  1. <p align="center">
  2. <img src=".github/assets/hero.png" alt="Attention Control: air traffic control discipline for agent output, written for a reader with ADHD. Two flight progress strips. The unstyled reply overruns a single box and falls off the edge. The Attention Control reply splits into four fields: ACTION, EDIT, STATE, NEXT." width="100%">
  3. </p>
  4. <p align="center">
  5. <strong>Attention Control</strong><br>
  6. <em>Air traffic control discipline for agent output.</em><br>
  7. <em>Written for a reader with ADHD.</em>
  8. </p>
  9. <p align="center">
  10. <a href="LICENSE"><img src="https://img.shields.io/github/license/aaddrick/attention-control?style=flat" alt="License"></a>
  11. <a href=".github/workflows/plugin-load-check.yml"><img src="https://img.shields.io/github/actions/workflow/status/aaddrick/attention-control/plugin-load-check.yml?label=plugin%20loads&style=flat" alt="Plugin load check"></a>
  12. </p>
  13. <p align="center">
  14. <a href="https://www.linkedin.com/in/aaddrick/">Connect on LinkedIn!</a>
  15. </p>
  16. <p align="center">
  17. <strong>English</strong> ·
  18. <a href=".github/readme/README.zh-CN.md">简体中文</a> ·
  19. <a href=".github/readme/README.ja.md">日本語</a> ·
  20. <a href=".github/readme/README.ko.md">한국어</a> ·
  21. <a href=".github/readme/README.vi.md">Tiếng Việt</a> ·
  22. <a href=".github/readme/README.pt-BR.md">Português (BR)</a>
  23. </p>
  24. ## Install
  25. <details open>
  26. <summary><strong>Claude Code</strong></summary>
  27. Claude Code is the only agent here with a native output style slot. Run these
  28. two commands in your terminal:
  29. ```bash
  30. claude plugin marketplace add aaddrick/attention-control
  31. ```
  32. ```bash
  33. claude plugin install attention-control@attention-control
  34. ```
  35. Then run `/config`, select **Output style**, and pick **Attention Control**. It
  36. takes effect after `/clear` or the next session.
  37. To skip the picker, add `outputStyle` to `~/.claude/settings.json`. It is a
  38. top-level key. It does not go inside `env`, `permissions`, or any other block:
  39. ```json
  40. {
  41. "model": "opus",
  42. "env": { "EXAMPLE_VAR": "1" },
  43. "outputStyle": "Attention Control"
  44. }
  45. ```
  46. `model` and `env` stand in for keys you may already have. Keep them. Add the
  47. `outputStyle` line beside them.
  48. For one session instead of every session, use the skill the plugin also ships:
  49. ```
  50. /attention-control:attention-control
  51. ```
  52. Say "stop attention control" to turn it off.
  53. </details>
  54. <details>
  55. <summary><strong>Codex</strong></summary>
  56. Codex has no output style slot, so the rules ship as a skill.
  57. ```bash
  58. codex plugin marketplace add aaddrick/attention-control --ref main
  59. ```
  60. ```bash
  61. codex plugin add attention-control@attention-control
  62. ```
  63. Inside Codex, `/plugins` opens the plugin browser instead.
  64. Start a new thread, then type the skill:
  65. ```
  66. $attention-control:attention-control
  67. ```
  68. Codex prefixes a plugin skill with the plugin name. Say "stop attention control"
  69. to turn it off. For every turn instead, put the always-on snippet from
  70. [INSTALL.md](./INSTALL.md#the-always-on-snippet) in `~/.codex/AGENTS.md`.
  71. </details>
  72. <details>
  73. <summary><strong>Cursor, Gemini CLI, Copilot, Zed, and manual install</strong></summary>
  74. See [INSTALL.md](./INSTALL.md). None of them has an output style slot, so the
  75. rules ship as a skill, a rules file, or an `AGENTS.md` block. On a skill route,
  76. type `/attention-control` and say "stop attention control" to turn it off.
  77. </details>
  78. ## What it does
  79. Air traffic control phraseology exists for one reason: a distracted human under
  80. load will mishear an instruction. Aviation solved this with two disciplines.
  81. Controlled vocabulary makes each word mean one thing. Fixed message shape puts
  82. the instruction first and the background last.
  83. This style applies both disciplines to your coding agent. The agent leads with
  84. the action you can run, and writes each sentence so one word means one thing.
  85. The style targets one reader: a reader with ADHD. That reader is where the shape
  86. rules come from. See [Why the shape rules exist](#why-the-shape-rules-exist).
  87. ## What changes
  88. <table>
  89. <tr>
  90. <td width="50%">
  91. ### Before
  92. > Great question! Let me take a look. It seems like the auth flow has a few
  93. > moving pieces here. Looking at `src/auth.ts`, the token verification logic
  94. > could possibly be utilizing a deprecated API — the `verifyToken` function
  95. > around lines 42-58 has been implemented against an older `jsonwebtoken`
  96. > signature. One approach that might be considered would be updating the
  97. > package and refactoring that function. By the way, your dependency versions
  98. > in general could probably use a review at some point. Hope this helps! Let me
  99. > know if you'd like me to dig deeper into any of this.
  100. </td>
  101. <td width="50%">
  102. ### After
  103. > Run `npm install jsonwebtoken@latest`.
  104. >
  105. > I changed `src/auth.ts:47`. `verifyToken` now names the algorithm:
  106. >
  107. > `jwt.verify(token, secret, { algorithms: ["HS256"] }, callback);`
  108. >
  109. > Run `npm test -- auth.spec.ts`. It takes about 10 seconds.
  110. >
  111. > The old call named no algorithm. `jsonwebtoken` v9 rejects an unsigned token
  112. > unless you pass `algorithms: ["none"]`.
  113. >
  114. > Next: paste the first failing line if a test fails.
  115. </td>
  116. </tr>
  117. </table>
  118. ## The two layers
  119. **Shape** decides what you say and in what order. 11 rules:
  120. 1. Lead with the next action.
  121. 2. Do the work you own.
  122. 3. Number multi-step work.
  123. 4. End with one concrete next action.
  124. 5. Suppress tangents.
  125. 6. Restate state every turn.
  126. 7. Give time estimates in concrete units.
  127. 8. Show what now works.
  128. 9. State errors flat.
  129. 10. Cap lists at 5 items.
  130. 11. No preamble, no recap, no closer.
  131. **Language** decides how each sentence reads. The short version:
  132. - One word, one meaning. One action, one verb. No rotating synonyms.
  133. - Standard verbs: "check", "make sure", "start", "stop", "use", "show", "find", "change", "remove", "need".
  134. - Active voice. Name the actor.
  135. - Simple tenses only. No perfect tense, no auxiliary stacks.
  136. - 20 words per instruction, 25 per explanation. Noun clusters capped at 3 words.
  137. Full text: [`output-styles/attention-control.md`](./output-styles/attention-control.md).
  138. ## Why the shape rules exist
  139. Five facts about ADHD reading drive all 11 shape rules. Each fact below names
  140. the rules it produces.
  141. | The fact | What the agent does |
  142. |---|---|
  143. | **Working memory is small.** Anything not on screen is gone. | It never writes "keep in mind X". It restates the state every turn: "Step 3 of 5 done: I changed the schema. Next: run `scripts/backfill.py`." (rules 6, 10) |
  144. | **Knowing the answer is not doing the answer.** Work dies in the gap between the two. | It does the work it owns instead of handing it back. It gives the command, not the label. "Add the missing header" is a label. `Authorization: Bearer ${token}` is a fix. (rules 1, 2, 3) |
  145. | **Starting is the hardest step.** | The first line is small, obvious, and doable now. The last line names one action that takes under two minutes. "Open the file" counts. (rules 1, 4) |
  146. | **Time estimates feel uniform.** "A bit of work" and "a few hours" register the same. | It writes "about 15 minutes if tests cover this, an afternoon if not". It never writes "some work". (rule 7) |
  147. | **Dopamine is scarce.** A buried win does not register. | After a change, it names the result in concrete terms: "Login works with magic links. Run `npm run dev` and open `/login`." (rule 8) |
  148. Two more rules protect attention itself. Rule 5 suppresses tangents, so one open
  149. thread stays one open thread. Rule 11 removes the preamble and the closer, so
  150. the answer starts on line 1.
  151. This is why the style is not "be terse". Terseness that drops the command, the
  152. number, or the condition costs the reader a round trip, and a round trip costs
  153. the thread. Rule 9 follows from the same logic: an error gets a location, a
  154. cause, and a fix, with no "Uh oh" in front of it. Alarm is not information, and
  155. it competes with the information for the same attention.
  156. You need no ADHD diagnosis for this to help. A tired reader, a reader on a
  157. phone, and a reader with 40 open tabs all read the same way.
  158. ## What it leaves alone
  159. Four targets, one rule each. Code, commands, file paths, identifiers, and error
  160. messages stay verbatim. Quoted text stays verbatim. Code comments and commit
  161. messages match the style of the repository around them. Only the prose the agent
  162. writes itself follows this style.
  163. Those are instructions, not a lock. An output style is system prompt text, and
  164. nothing outside the model enforces it. Treat the boundary as a strong default,
  165. and check the output when it matters.
  166. Accuracy beats brevity. A rule never removes a fact, a number, a condition, or a
  167. scope qualifier. A hedge that carries real uncertainty stays.
  168. ## Evals
  169. The harness compares response quality against an unstyled baseline. It does not
  170. measure length.
  171. ```bash
  172. python3 scripts/run_evals.py validate
  173. ```
  174. ```bash
  175. python3 scripts/run_evals.py plan --trials 3
  176. ```
  177. 24 cases, 6 scored dimensions, and a release gate that blocks a candidate that
  178. regresses correctness or safety.
  179. The judge is the weak point, so the harness targets it. `blind` hides the
  180. condition and balances the positions. The judge scores every group twice with
  181. the order reversed, then reports how often the two passes disagree. The runner
  182. runs in an empty directory and reads none of your config. Design notes and the
  183. measurements behind them: [evals/README.md](./evals/README.md).
  184. ## Tune it
  185. Fork, edit `output-styles/attention-control.md`, then regenerate every
  186. agent-specific copy:
  187. ```bash
  188. python3 scripts/sync_style.py
  189. ```
  190. Swap your copy in, one command at a time:
  191. ```bash
  192. claude plugin uninstall attention-control
  193. ```
  194. ```bash
  195. claude plugin marketplace remove attention-control
  196. ```
  197. ```bash
  198. claude plugin marketplace add <your-username>/attention-control
  199. ```
  200. ```bash
  201. claude plugin install attention-control@attention-control
  202. ```
  203. ## Credits
  204. This style combines two existing works. Neither author takes part in this
  205. project.
  206. **Shape layer:** [`i-have-adhd`](https://github.com/ayghri/i-have-adhd) by
  207. Ayoub G. (MIT). The eval harness derives from the same project.
  208. **Language layer:** the
  209. [`asd-ste100` output style](https://gist.github.com/L1nefeed/4164ecaaf77879e76dca3c06f142f1c2)
  210. by [L1nefeed](https://github.com/L1nefeed), itself a condensation of
  211. [ASD-STE100](https://www.asd-ste100.org/) Simplified Technical English, Issue 9.
  212. This project reproduces no text from the ASD specification. The ASD does not
  213. certify, endorse, or sponsor it. Details in [NOTICE.md](./NOTICE.md).
  214. ## License
  215. MIT. See [LICENSE](./LICENSE).

Ratings and comments

Sign in to rate this skill.

Sign in

Free to use. If it’s helped, pass it on.