ste80

agent
Security Audit
Warn
Health Warn
  • License — License: MIT
  • Description — Repository has a description
  • Active repo — Last push 0 days ago
  • Low visibility — Only 5 GitHub stars
Code Pass
  • Code scan — Scanned 12 files during light audit, no dangerous patterns found
Permissions Pass
  • Permissions — No dangerous permissions requested

No AI report is available for this listing yet.

SUMMARY

STE80 Humanizer: agent skill

README.md

STE80

CI License: MIT

A Claude Code plugin with one skill, ste80. The skill makes AI-written text read like a careful person wrote it. It does not change what the text says. It works in two passes:

  1. Remove the AI tells. The 26 tells from Wikipedia's Signs of AI writing, in the catalog of blader/humanizer.
  2. Rebuild the sentences in STE80. That is "80% of the way to ASD-STE100", the method that Andrej Karpathy shared. ASD-STE100 is the controlled English of aircraft maintenance manuals.

The skill text comes from Dodaios/ste80-humanizer, with small changes. This version needs no Python. Its checker is a POSIX sh and awk script.

Before:

Great question! Kubernetes liveness probes are a powerful mechanism that enables the kubelet to determine whether a container is healthy — and it's not just a health check, it's your first line of defense. When a probe fails, the container will be automatically restarted by the kubelet, ensuring that your application remains available. However, it's important to note that misconfigured probes could potentially cause cascading restarts, so parameters like initialDelaySeconds and failureThreshold should be carefully tuned.

Get this right, and your cluster heals itself.

After:

A liveness probe tells the kubelet if a container is healthy. If the probe fails, the kubelet restarts the container. This keeps the application available.

But a misconfigured probe can cause cascading restarts. To prevent this, tune parameters like initialDelaySeconds and failureThreshold with care.

The goal is text that people read easily. The skill does not try to get past AI detectors.

Install

claude plugin marketplace add adrianhaj/ste80
claude plugin install ste80@ste80

To update to a newer release, refresh the marketplace, update the plugin and restart Claude Code:

claude plugin marketplace update ste80
claude plugin update ste80@ste80

For another agent, copy the folder skills/ste80 into the skills folder of that agent. The folder is a plain agent skill: one SKILL.md, its references and the checker.

Use

/ste80:ste80
<paste text>

Or ask in your own words:

This sounds like ChatGPT wrote it. Make it sound like a person.
Rewrite this README 80% of the way to ASD-STE100.
Explain how TLS handshakes work in STE.
Light touch: just remove the AI stuff from docs/intro.md.
Convert docs/runbook.md to strict STE.
Does this email sound like AI? Just tell me, don't rewrite it.
Mode When What it does
Humanize (default) Normal requests Pass 1 fully, Pass 2 at 80%. Keeps your voice.
Light "light touch", "just remove the AI stuff" Pass 1, plus only the worst sentences from Pass 2.
Strict STE "strict", "full STE", procedures, safety notes Pass 1 and all of Pass 2 at 100%.

The skill keeps every fact. It keeps "may" as "may" and "40+" as "40+". It adds nothing, and it flags contradictions in the source. If you add a sample of your own writing, the skill matches its sentence length, words and punctuation.

The method

Andrej Karpathy suggested that you ask an LLM to write "80% of the way to ASD-STE100", because the full specification is too strict for normal prose.

ASD-STE100 (Simplified Technical English) is from the aerospace industry of the 1980s. Its writers made it for mechanics whose first language is not English. Its rules:

  • 20 words per instruction and 25 per description
  • one instruction per sentence
  • active voice and simple tenses
  • one meaning per word, with about 900 approved words
  • one topic per paragraph, and six sentences at most

At 80%, the skill keeps all of the structure and only as much of the vocabulary as still sounds natural:

STE rule 100% (Strict mode) 80% (default)
Sentence length 20 / 25 Hard limit A few words over is fine
One idea per sentence Always Always
Active voice, simple tenses Always Always
Approved dictionary only Yes No, use the plain common word
"check", "test" as verbs No Yes
"may", "should" No Yes, for real uncertainty
Contractions No Yes, in casual text
Semicolons No No

STE fixes sentence structure. It does not remove chatbot phrases, so Pass 1 does that first.

Layout

Path What
skills/ste80/SKILL.md The skill: modes, the 26 AI tells (§1 to §26), the 15 STE80 rules (S1 to S15), fidelity rules
skills/ste80/references/ai-tells.md Each tell in full: Watch for, Problem, Before, After
skills/ste80/references/ste80-rules.md Examples for each STE80 rule
skills/ste80/references/strict-mode.md The strict rule set and a checklist
skills/ste80/references/word-swaps.md More than 100 plain-word swaps
skills/ste80/references/worked-example.md One text after Pass 1 and after Pass 2
skills/ste80/scripts/ste_check.sh, ste_check.awk The checker
evals/ Eight cases in the claude plugin eval format
tests/ Tests for the checker

The skill reports its edits by number, for example "§1, §8, S2", so you can look each one up.

Checker

skills/ste80/scripts/ste_check.sh is a small linter. It needs only sh and awk, which every Unix system has. On Windows, use Git Bash or WSL.

sh skills/ste80/scripts/ste_check.sh README.md
sh skills/ste80/scripts/ste_check.sh --strict docs/runbook.md
cat draft.txt | sh skills/ste80/scripts/ste_check.sh -
Pass What the checker flags
1 not-X-but-Y, clipped tails, rows of fragments, closers, chatbot phrases, run-ups, fake depth, stacked hedges, vague authority and association, "-ing" riders, inflated claims, AI words, repeated openings, hyphen pairs, bold labels, Title Case and emoji in headings, curly quotes, leftovers from a chatbot (link tags, placeholders, citation marks)
2 sentence and paragraph length, sentences of the same length in a row, semicolons, dashes, passive voice, perfect and progressive tenses, "-ing" forms, phrasal verbs

The exit code is 0 for no findings, 1 for findings and 2 for a usage error. The checker is a heuristic. It is not an AI detector and it is not a compliance checker.

Develop

sh tests/test_ste_check.sh
shellcheck skills/ste80/scripts/ste_check.sh tests/test_ste_check.sh
claude plugin validate .

CI runs the tests with gawk, mawk, busybox awk and the BWK awk that macOS uses. It also runs them in Git Bash on Windows. To test one awk yourself, set STE_AWK, for example STE_AWK=mawk sh tests/test_ste_check.sh.

tests/expected/ holds the output that the checker must give for tests/fixtures/sample.md. If you change a rule on purpose, read the new output, then write it to those files.

To run the eval cases:

claude plugin eval . --allow-tools Write,Edit,Bash

The run does each case with the plugin and without it. Each case has a grader 00-skill-loaded. It shows if the skill loaded, and it does not count in the score. If the skill did not load, the score of that case says nothing about the skill.

Rules for a change to the skill:

  • Write in the style that the skill teaches.
  • Keep SKILL.md below 500 lines. Long examples go in references/.
  • Keep the checker in sync with the tells and words in SKILL.md, where a pattern can find them.
  • Do not copy the ASD-STE100 specification or its dictionary. ASD owns the copyright. Write the rules in your own words.
  • Do not claim compliance. The skill and the checker give "STE style", not certified STE.

Release

  1. Bump version in .claude-plugin/plugin.json as the last commit of the change (chore: release X.Y.Z). Users get the new version once that commit is on main.
  2. After the merge, .github/workflows/release.yml creates the tag ste80--vX.Y.Z on that commit. It also publishes a GitHub release that lists the pull requests since the previous release. It does nothing when that version already has a release.
  3. Optional: edit the release on GitHub to add a few lines about what changed above that list.

Only the code owner in .github/CODEOWNERS can merge to main, so only the code owner can start a release.

Limits

  • STE is for clear explanatory text. For fiction, speeches and brand copy, the skill does Pass 1 only.
  • STE is an English standard. For other languages, the skill applies the tells and the structure rules, but not the English word lists.
  • "STE style" is not "STE compliant". Compliance needs the official dictionary, trained writers and review.

Credits

Both license notices are in THIRD_PARTY_NOTICES.md.

Disclaimer

This skill does not contain the ASD-STE100 specification or its dictionary. The STE rules are paraphrased and simplified. ASD-STE100 and Simplified Technical English are trademarks of ASD (AeroSpace, Security and Defence Industries Association of Europe). This project is not affiliated with or endorsed by ASD.

License

MIT

Reviews (0)

No results found