Skip to content
Back to search

nopus

Deterministic prose checks for clearer coding-agent responses

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/vistyy--nopus/badge.svg)](/listings/vistyy--nopus)

HTML

<a href="/listings/vistyy--nopus"><img src="/api/listings/vistyy--nopus/badge.svg" alt="findsafeskills safety rating" /></a>

Install

i

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 · 15 lines

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

  1. {
  2. "name": "nopus",
  3. "owner": {
  4. "name": "Vistyy"
  5. },
  6. "description": "Uses deterministic prose checks to flag overly complex coding-agent responses and ask for one clearer rewrite.",
  7. "plugins": [
  8. {
  9. "name": "nopus",
  10. "source": "./plugin",
  11. "description": "Uses deterministic prose checks to flag overly complex coding-agent responses and ask for one clearer rewrite.",
  12. "version": "1.1.0"
  13. }
  14. ]
  15. }
README.md · README · 277 lines

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

  1. ![nopus banner showing a speech bubble compressed into one clear word](https://raw.githubusercontent.com/Vistyy/nopus/main/assets/nopus-banner.png)
  2. # nopus
  3. [![npm version](https://img.shields.io/npm/v/%40syzom%2Fnopus)](https://www.npmjs.com/package/@syzom/nopus)
  4. [![Node.js version](https://img.shields.io/node/v/%40syzom%2Fnopus)](package.json)
  5. [![MIT license](https://img.shields.io/npm/l/%40syzom%2Fnopus)](LICENSE)
  6. nopus catches coding-agent answers that disappear into abstract LLM babble.
  7. Using deterministic prose checks, it flags responses that cross your chosen complexity threshold and sends them back for one clearer rewrite.
  8. ## See it in action
  9. ![A real coding-agent response, the prose issues caught by nopus, and the clearer rewrite](https://raw.githubusercontent.com/Vistyy/nopus/main/assets/nopus-rewrite-example.png)
  10. Currently supported: Pi, Claude Code, Codex
  11. ## Use nopus
  12. To use nopus, install Node.js 22 or newer and make sure the `node` command is available.
  13. <details>
  14. <summary>Pi</summary>
  15. Install nopus from npm:
  16. ```sh
  17. pi install npm:@syzom/nopus
  18. ```
  19. You can install the GitHub repository instead:
  20. ```sh
  21. pi install git:github.com/Vistyy/nopus
  22. ```
  23. Start Pi normally after installation.
  24. The extension checks each completed response.
  25. When it requests a rewrite, it hides the rejected response from the terminal transcript by default.
  26. The response remains in the Pi session and model history.
  27. Because nopus evaluates completed responses, the response remains visible while it streams and disappears only after nopus requests a rewrite.
  28. Use these commands while working:
  29. ```text
  30. /nopus status
  31. /nopus check
  32. /nopus on
  33. /nopus off
  34. /nopus extra-simple on
  35. /nopus extra-simple off
  36. /nopus hide-original on
  37. /nopus hide-original off
  38. ```
  39. </details>
  40. <details>
  41. <summary>Claude Code</summary>
  42. Add the marketplace from GitHub and install the plugin:
  43. ```text
  44. /plugin marketplace add Vistyy/nopus
  45. /plugin install nopus@nopus
  46. ```
  47. The plugin uses a Stop hook to request one clearer response.
  48. The identifier follows the `plugin@marketplace` format, so this plugin is `nopus@nopus`.
  49. </details>
  50. <details>
  51. <summary>Codex</summary>
  52. Add the marketplace from GitHub and install the plugin:
  53. ```sh
  54. codex plugin marketplace add Vistyy/nopus
  55. codex plugin add nopus@nopus
  56. ```
  57. Start a new Codex session after installation so the plugin and its skills load.
  58. The Codex plugin uses the same bounded Stop hook.
  59. The identifier follows the `plugin@marketplace` format, so this plugin is `nopus@nopus`.
  60. </details>
  61. ## How it works
  62. nopus examines the prose when a coding agent completes a response.
  63. Its measurements are deterministic, so the same prose and sensitivity produce the same decision.
  64. It measures these characteristics:
  65. - **Uncommon wording** - Words that are rare in ordinary conversation.
  66. - *Example:* “Commence the task” instead of “Start the task.”
  67. - **Very uncommon wording** - Words that are rare in both conversation and broad written English.
  68. - *Example:* “Ameliorate the problem” instead of “Fix the problem.”
  69. - **Abstract vocabulary** - Concepts and qualities used without concrete people, objects, or actions.
  70. - *Example:* “capability, strategy, governance, and ownership.”
  71. - **Abstract sentences** - Abstract wording that fills one sentence and continues across several sentences.
  72. - *Example:* “The strategy establishes capability ownership through governance transformation.”
  73. - **Noun and modifier stacks** - Three or more nouns or modifiers compressed into one phrase.
  74. - *Example:* “repository state mutation verification strategy.”
  75. - **Phrase load** - Dense phrases that appear frequently for the amount of prose.
  76. - *Example:* A short answer containing both “deployment policy validation process” and “repository state mutation strategy.”
  77. - **Formulaic style cues** - Literal phrases that add inflated framing or filler.
  78. - *Example:* “Here’s where it gets interesting.”
  79. Word rarity comes from packaged conversational and broad-web frequency data.
  80. Abstractness comes from human-rated word concreteness, and technical exceptions come from a packaged computing glossary.
  81. The style matcher uses only a small allowlist of literal cues rather than trying to classify arbitrary writing habits.
  82. nopus then looks for combinations that make a response materially harder to read.
  83. For example, uncommon words become more significant when they appear with sustained abstraction and dense phrase stacks.
  84. Several abstract sentences become more significant when loaded phrases recur throughout them.
  85. ### How nopus avoids unnecessary rewrites
  86. - It removes code blocks, inline code, URLs, file paths, and table rows before measuring prose.
  87. - It excludes obvious identifiers and reduces the weight of established computing terms.
  88. - It uses proportions and phrase density, so a long response is not penalized merely for being long.
  89. - Most decision paths require several measurements to cross their thresholds together.
  90. - A single rare word or dense phrase does not normally trigger a rewrite.
  91. - Style cues require either two distinct cues or one cue supported by enough uncommon wording.
  92. - It requests at most one automatic rewrite, so it cannot create a retry loop.
  93. Necessary technical terms, commands, identifiers, conditions, and qualifications remain part of the requested answer.
  94. `InteractiveSessionHost` stays `InteractiveSessionHost`; nopus targets the difficult prose around it.
  95. When nopus requests a rewrite, the host displays this message:
  96. ```text
  97. nopus requested a clearer rewrite.
  98. ```
  99. ### What the agent receives
  100. For the response used in the example below, the default rewrite request gives the agent focused guidance and examples from its own response:
  101. ```text
  102. Rewrite the response for clarity and directness.
  103. Keep its meaning and necessary detail.
  104. Return only the revised response.
  105. Focus on these areas:
  106. - Replace formulaic framing with plain statements.
  107. Examples from the response:
  108. - "here's where it gets interesting": lecture-hall framing.
  109. - "paradigm shift": tired.
  110. ```
  111. Turning rewrite evidence off removes the `Examples from the response` block.
  112. The agent still receives the rewrite instructions and focus areas.
  113. Extra-simple rewriting is off by default.
  114. When enabled, nopus requests a short initial answer that keeps the main conclusion, immediate action, and conditions or warnings that could change that action.
  115. It leaves supporting rationale, examples, alternatives, future considerations, and lower-priority implementation details for follow-up.
  116. ## Choose how sensitive it should be
  117. Sensitivity controls how often nopus intervenes, and the default is `medium`.
  118. | Sensitivity | What to expect | Observed rewrite rate | Best fit |
  119. |---|---|---:|---|
  120. | `low` | Uses the strictest thresholds to target clear cases where abstraction, uncommon wording, or dense phrasing accumulates strongly. | 5.3% | You want intervention only for the most difficult responses. |
  121. | `medium` | Catches difficulty sustained across several sentences or supported by multiple characteristics without treating isolated wording as a problem. | 9.9% | You want a balanced default for regular use. |
  122. | `high` | Uses lower thresholds so shorter runs of abstraction and borderline combinations can request a rewrite sooner. | 18.6% | You consistently prefer plain language and accept more interventions. |
  123. The [observed rates](evaluation/README.md) come from 5,337 completed Pi responses and provide only a rough comparison because results vary by agent and task.
  124. ## Configure automatic rewrites
  125. `nopus-configure` changes the settings used by the automatic integrations.
  126. It does not rewrite a response.
  127. | Action | Pi | Claude Code | Codex |
  128. |---|---|---|---|
  129. | Set medium sensitivity | `/skill:nopus-configure medium` | `/nopus:nopus-configure medium` | `$nopus-configure medium` |
  130. | Disable rewrite evidence | `/skill:nopus-configure evidence off` | `/nopus:nopus-configure evidence off` | `$nopus-configure evidence off` |
  131. | Enable extra-simple rewrites | `/skill:nopus-configure extra-simple on` | `/nopus:nopus-configure extra-simple on` | `$nopus-configure extra-simple on` |
  132. | Show rejected Pi responses | `/skill:nopus-configure hide-original off` | - | - |
  133. <details>
  134. <summary>Configuration file and environment variables</summary>
  135. The configuration skill writes `$XDG_CONFIG_HOME/nopus/config.json`, or the platform user-configuration equivalent.
  136. ```json
  137. {
  138. "complexitySensitivity": "medium",
  139. "includeEvidence": true,
  140. "extraSimple": false,
  141. "pi": {
  142. "hideOriginalResponse": true
  143. }
  144. }
  145. ```
  146. Rewrite evidence and Pi response hiding are enabled by default.
  147. Extra-simple rewriting is disabled by default.
  148. `NOPUS_COMPLEXITY_SENSITIVITY`, `NOPUS_INCLUDE_EVIDENCE`, `NOPUS_EXTRA_SIMPLE`, and `NOPUS_PI_HIDE_ORIGINAL_RESPONSE` override the configuration file for automation.
  149. </details>
  150. ## Simplify a response manually
  151. `nopus-simplify` is the manual counterpart to the automatic rewrite.
  152. Use it when nopus does not trigger but you still find the preceding response too complex.
  153. Invoking the skill requests a rewrite immediately instead of applying the selected sensitivity threshold.
  154. It uses the same clarity goals as the automatic rewrite and preserves meaning, necessary detail, technical terms, commands, identifiers, conditions, and qualifications.
  155. It does not change nopus configuration.
  156. | Pi | Claude Code | Codex |
  157. |---|---|---|
  158. | `/skill:nopus-simplify` | `/nopus:nopus-simplify` | `$nopus-simplify` |
  159. ## Development
  160. Install dependencies and run all checks:
  161. ```sh
  162. pnpm install
  163. pnpm check
  164. ```
  165. Update the normalized datasets from their checksum-pinned upstream sources:
  166. ```sh
  167. pnpm import:data
  168. ```
  169. See [`evaluation/README.md`](evaluation/README.md) for the private-corpus calibration and regression workflow.
  170. Normal analysis and verification do not download runtime data.
  171. <details>
  172. <summary>Data sources and attribution</summary>
  173. nopus retains normalized JSON rather than the original upstream formats.
  174. The SHA-256 values below identify the exact upstream copies used to produce the normalized datasets.
  175. ### Claudisms
  176. `data/ai-style-cues.json` is derived from <https://claudisms.ai/claudisms.md>.
  177. The source describes itself as a living list of AI-writing habits that is free to copy and adapt.
  178. The copy was retrieved on 2026-08-14 and has SHA-256 `f4a09fa849b61712838d3e56c854ddf37a59417823acb348a330721615f73b2b`.
  179. ### SUBTLEX-US word frequencies
  180. `data/conversational-word-frequencies.json` is derived from `subtlex-word-frequencies@2.0.0`, available at <https://github.com/words/subtlex-word-frequencies>.
  181. The package derives its data from SUBTLEX-US by Marc Brysbaert and Boris New.
  182. The source paper is “Moving beyond Kučera and Francis: A Critical Evaluation of Current Word Frequency Norms and the Introduction of a New and Improved Word Frequency Measure for American English,” DOI <https://doi.org/10.3758/BRM.41.4.977>.
  183. The source copy has SHA-256 `271c5a5fbf332f60762cfa34b11394427c220099d96c589751b6bc77e5b32c1a`.
  184. The npm package declares the ISC license, which is preserved at `data/licenses/conversational-word-frequencies-ISC.txt`.
  185. ### Norvig web word counts
  186. `data/broad-web-word-counts.json` is derived from <https://norvig.com/ngrams/count_1w.txt>.
  187. Peter Norvig publishes the file as part of the data described at <https://norvig.com/ngrams/>.
  188. The counts derive from the Google Web Trillion Word Corpus described by Thorsten Brants and Alex Franz and distributed by the Linguistic Data Consortium.
  189. The source copy has SHA-256 `51df159fd3de12b20e403c108f526e96dbd723d9cabdd5f17955cdc16059e690`.
  190. ### Brysbaert concreteness ratings
  191. `data/concreteness-ratings.json` is derived from the ArtsEngine mirror at <https://github.com/ArtsEngine/concreteness>.
  192. The data accompanies Marc Brysbaert, Amy Beth Warriner, and Victor Kuperman, “Concreteness ratings for 40 thousand generally known English word lemmas,” DOI <https://doi.org/10.3758/s13428-013-0403-5>.
  193. The source copy has SHA-256 `0b4082dbd38585b0ee1fd258145b7a50592f8d0d98e5fc6b6844ceef3cd8ecc8`.
  194. ### Carpentries Glosario
  195. `data/technical-terms.json` is derived from The Carpentries `glosario` repository at revision `a4a774b6dea3881a7794d63aee10fbb5d34321d1`.
  196. The source repository is <https://github.com/carpentries/glosario>, and its archived releases are available under DOI <https://doi.org/10.5281/zenodo.13589476>.
  197. The source copy has SHA-256 `93a6371add7275d643cacb060ae0b65749d1ffdff329064feeaeaf7152c3715f`.
  198. The data is licensed under Creative Commons Attribution 4.0 International, and the license is preserved at `data/licenses/computing-glossary-CC-BY-4.0.md`.
  199. The Carpentries does not endorse nopus.
  200. </details>

Ratings and comments

Sign in to rate this skill.

Sign in

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