editorial-humanization
Skill0 starsChanged 55 days agoMITView on GitHub
Meaning-preserving editing of course text to read naturally and clearly (readability, flow, removing machine-sounding templating, polarity and modality preservation). Used by CourseForge EDITORIAL edits and editorial reviewers.
In this repository19 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- accessibility-reviewSkill
- alignment-reviewSkill
- assessment-designSkill
- assessment-reviewSkill
- course-gap-analysisSkill
- fact-checkSkill
- html-intake-reviewSkill
- instructional-designSkill
- instructional-graphicsSkill
- research-authoringSkill
- research-planningSkill
- research-reviewSkill
- scenario-designSkill
- storyboard-authoringSkill
- ui-reviewSkill
- ux-reviewSkill
- visual-directionSkill
The file
We split this skill into the rules it gives an agent
4 rules. About 0 are common, 3 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈3
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 37
Modality: must/shall/is required (duty) ≠ should/is recommended (guidance) ≠ may/can (option or possibility).
- Distinctive: line 47
Read the whole module before editing anything.
- Distinctive: line 50
Keep edits proportionate: fix the sentence, not the whole paragraph, unless the paragraph is the problem.
- Distinctive: line 51
When the best fix would change meaning (for example the text is wrong), do not make it.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Editorial humanisation
The one rule
Change how things are said, never what is said. A reader of the before and after texts must come away believing exactly the same things, with the same confidence, about the same scope.
What makes course text sound machine-written
- The same opener on block after block: "In this section…", "It is important to note that…", "Let's explore…".
- Filler intensifiers and buzzwords: crucial, vital, robust, seamless, delve, landscape, navigate the complexities, in today's fast-paced world, plays a pivotal role.
- Reflexive triads ("clear, concise, and compelling") and paired antitheses used as decoration.
- Summaries that restate the heading, and conclusions that begin "In conclusion" or "Overall".
- Stacked rhetorical questions opening every topic.
- Formulaic feedback ("Correct! Great job!", "Not quite. Try again!").
- Uniform sentence length and rhythm across a whole module.
- Hedging piled on hedging ("may potentially help to possibly reduce").
- Promotional tone for neutral material.
What good course prose does
- Leads with the point; one idea per sentence where the idea is hard.
- Uses the actor as subject when duties matter: "The employer must…" rather than "It must be ensured that…".
- Chooses concrete verbs and nouns; keeps the defined technical term rather than rotating synonyms.
- Varies sentence length naturally; uses lists for steps and criteria.
- Speaks to the learner as a capable adult: direct, warm, never cute.
- Connects blocks with meaningful transitions ("That is why the label alone is not enough; the next question is what your task does to exposure.").
Meaning you must preserve
- Polarity and contrast. Keep every negation and contrast in meaning: not, no, never, none, unless, except, only, not only… but also, rather than, instead of, without, neither/nor, cannot, alone, but. The classic failure: "not only the product classification" rewritten as "the product classification alone", which reverses the point. When you restructure a sentence, re-read it for polarity before keeping it.
- Modality. must/shall/is required (duty) ≠ should/is recommended (guidance) ≠ may/can (option or possibility). Never strengthen or soften.
- Scope. some/most/all; in a named jurisdiction; for listed substances; where reasonably practicable; at the time of writing. Never drop or broaden.
- Numbers and units. Unchanged, including ranges and "at least" or "up to".
- Causality and sequence. because, therefore, before, after, until; list order that encodes a process.
- Assessment meaning. The stem asks the same thing; each option means the same thing; the key stays the only correct answer; each distractor stays wrong for the same reason; feedback matches the key.
- Safety and escalation. "Stop and ask", "do not attempt", "involve a qualified…" keep their force.
- Terms. Hazard is not risk; correction is not corrective action; a defined term is never swapped for a near-synonym.
Editing practice
- Read the whole module before editing anything; note patterns that repeat across blocks.
- Edit only where there is a clear gain. Good text stays as it is.
- For each changed field, reread before and after side by side and check polarity, modality, scope, numbers and terms.
- Keep edits proportionate: fix the sentence, not the whole paragraph, unless the paragraph is the problem.
- When the best fix would change meaning (for example the text is wrong), do not make it; note it for a human.
Reviewing an editorial pass
Compare each edit with its original. Report reversed or changed meaning as critical, dropped qualifiers or modality shifts as major, and edits that made text worse as minor. Quote before and after text as evidence. Code already verifies structural identity (IDs, keys, citations, numbers, objective maps) and flags removed negation markers; the reviewer judges logic that changed without a marker word disappearing.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

