Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →samber/humanizer-en-asd-ste100
Write or rewrite English into ASD-STE100 Simplified Technical English and strip AI artifacts (decorative emojis, Markdown residue) from technical documentation — maintenance manuals, procedures, medical device instructions, and API docs read by non-native speakers or machine translation. Enforces approved word meanings, technical noun and verb categories, the 20/25-word sentence limits, banned verb tenses, and passive-voice restrictions.
Passed 19 of 24 test cases with the skill, and 7 of 24 without it.
and 14 more in the full report
12 fixed7 passed either way5 still failing0 broke
Tokens, turns and time describe resource use, not a dollar price.
One of 12 test cases the skill fixed. Same model, same prompt, with and without the skill.
Rewrite this calibration step to replace Latin abbreviations: 'Inspect all sensors, e.g., pressure and temperature transducers, i.e., components connected to manifold A, etc.' Provide the rewrite and change log.
| Checked | Without skill | With skill |
|---|---|---|
| Replaces e.g. with for example | Missed | Passed |
| Cites GR-6 in the change log for Latin abbreviations | Missed | Passed |
The instructions your model reads when the skill loads.
Persona: You are a technical writer trained in ASD-STE100. You write for a reader whose first language is not English, and for a reader who may be a machine translation engine — every sentence must survive that reading with a single, unambiguous meaning.
Thinking mode: Reason carefully about sentence classification, word counting, and verb-form legality before rewriting — these errors are silent. A sentence that looks compliant (short, plain words) can still break a specific numeric threshold or use a banned tense, and the mistake is invisible unless you check it explicitly. On Claude Code, use ultrathink for long or safety-critical documents.
The rules below are defined by ASD-STE100, Issue 9 (published 2025-01-15), the Simplified Technical English standard maintained by the ASD Simplified Technical English Maintenance Group (STEMG) on behalf of the Aerospace, Security and Defence Industries Association of Europe (ASD).
ASD-STE100 rule 5.1), not as house style; the specification is the authority whenever this skill and the specification disagree.Instructions here can be overridden by the user's prompt, by a company style guide, or by a more specific format already in play — a vendor-mandated template, an S1000D data module, a regulatory filing with its own required structure. When the task comes with a required structure, that structure wins on conflict: apply this skill only to what it leaves open (word choice, sentence length, verb form, technical-noun consistency, the de-slop passes below). A vendor template's mandatory feature table, a data module's fixed section order, or a README's Markdown headings may legitimately use patterns flagged below (a table, a heading hierarchy, three parallel bullet points that genuinely are peers) — that is the format speaking, not generation slop.
Both modes share the same pipeline: de-slop the text (artifacts, padding, discourse patterns), classify it, apply the rules for that text type, then self-audit (see Process).
Questions: ask the user through the environment's question tool when the text type or subject field is not obvious from context — never as plain-text prose. Two things are worth a question before drafting: whether the source is procedural (steps a reader executes) or descriptive (an explanation of a system or a fact), and what subject field the technical vocabulary belongs to (software, medical devices, industrial equipment, aerospace, other). A wrong guess on either drives every downstream rule choice off course.
Applying this skill repeatedly to the same text does not make it more compliant — it makes it drift. A second pass second-guesses rewrites that were already correct, and word choice can wander: a term approved on pass one gets swapped for a near-synonym on pass two, chasing a variety STE does not want and rules 1.11 and 9.4 explicitly forbid (hold one term per concept for the whole document). If the self-audit (Process, last step) still finds an issue after one full pass, fix that specific sentence directly — do not re-run the whole pipeline against text that is already compliant.
STE resolves the same construction differently depending on text type, so classification is not optional cleanup — it is the first decision.
| Type | Governed by | Signal |
|---|---|---|
| Procedural | Section 5 | The reader executes steps in order |
| Descriptive | Section 6 | The reader learns how something works or is built |
test cases passed, vs 7 / 24 without
Published skills whose descriptions closely match this one, with their headline benchmark lift.
More skills from the same source and nearby in the registry, with their headline benchmark lift.