Search the docs, videos, blog, apps, teams, skills and release notes.
At a glance
What it is
Also valid as a one-step on a draft you know is AI-flavored and just wants fixed.
A skill is a written procedure a bot loads when a job calls for it. This one is a single file, SKILL.md, and the whole file is on this page.
Each of these library teams gives its bots this skill. Import one and look inside first.
The whole file, exactly as it downloads. Read it before you rely on it.
SKILL.md
112 lines · 951 words
Mode skill. Default-enabled on the Humanizer specialist.
Use when the user wants a draft rewritten to read human, usually after a diagnose pass surfaced specific tells. Also valid as a one-step on a draft the user knows is AI-flavored and just wants fixed.
Trigger phrases:
"Humanize this."
Also valid as a one-step on a draft you know is AI-flavored and just wants fixed.
Nothing. It ships with Murage, which is open source. Your bots run on the AI plan you already have.
Yes. Open it in Settings → Skills to read, edit or duplicate it. Skill Guard scans it again after every edit.
Download the app, connect the AI you already pay for, and tell your Chief of Staff what needs doing. Plan on about ten minutes from install to a working team.
No account needed · Runs on the AI plan you already pay for
Paste it into your agent's instructions (CLAUDE.md, AGENTS.md or a custom GPT), or add it to a bot in Murage.
Ask Ember for the bot, then switch the skill on in its settings.
Step 1
Send your Chief of Staff this message. It proposes the bot and its role first, and creates it when you say go.
Send this to Ember
“Create a bot that uses the Rewrite Pass skill, and show me the plan first.”
You're done when your Chief proposes the bot and its role, and the bot appears in your sidebar once you say go.
Step 2
You're done when you see the bot’s skills and the library.
Step 3
You're done when the skill shows as on for that bot.
Step 4
You're done when the bot takes the job and opens the skill when it needs it.
Prefer to do it yourself? Choose + in the sidebar, then New Bot.Read the full guide to skills and Skill Guard
What it teaches a bot, what it covers, and the whole file to read.
The principle
**Mode skill.** Default-enabled on the Humanizer specialist.
What it covers
When to use
Procedure
Decision rules
Anti-patterns
Before / after
"Make this sound like a person wrote it."
"Run a humanize pass on this draft."
If the draft hasn't been diagnosed yet, run a quick mental diagnosis first (AI-flatness score, rhythm score, top tells) so the rewrite is targeted. Do not dump the full diagnostic on the user unless asked. The rewrite output is the deliverable.
1. Substitute vocabulary. Replace every blacklist word with a human-register alternative. See humanizer-tell-detector for the canonical list. Context-shifts, not 1-to-1 synonyms:
delve / dive deep → dig into, look at
leverage → use, lean on
navigate → work through, deal with
pivotal → important, load-bearing
robust → strong, solid
multifaceted / intricate → complicated, layered
comprehensive → full, covers everything
streamline → tighten, cut
"in today's digital age" → today, in 2026
"Furthermore / Moreover / Additionally" → cut, or "Also"
When a word has no clean substitute, rewrite the sentence to remove the need.
2. Vary sentence length (Klinkenborg's rule). Never two consecutive sentences within 3 words of each other in length. Target rhythm patterns:
Short / long / short, the most reliable humanizing rhythm
Long / fragment / long. Fragments are humanizing, not errors
A one-word sentence somewhere in every 200 words
If the input has five 15-word sentences in a row, the output should have lengths like 7 / 24 / 11 / 3 / 19. Rewrite to hit the rhythm even if it means restructuring the paragraph.
3. Break parallel structure. AI loves parallelism: "She did A, did B, did C." "It is X, it is Y, it is Z." "First... Second... Finally..." When you see three or more parallel clauses or sentences, break two of them into non-parallel constructions.
4. Contraction inconsistency. AI over-expands contractions. Convert about 60% of do not / it is / we are / cannot / will not / would not / should not to contractions and leave the rest expanded. The mixing itself is a strong human signal; uniform expansion or uniform contraction both read flat.
5. One idiom or colloquialism per ~250 words. AI under-uses these. Starter list: "hit the ground running," "the catch is," "long story short," "the elephant in the room," "moving the needle," "on the back of," "for what it is worth," "that said." Use one. Do not stack three; cliché-pile reads worse than the original.
6. Add a parenthetical or aside. In every 300 words of rewritten text, include at least one personal aside, parenthetical, or hedge: "(or you can," "in my experience," "honestly," "though that depends." This is the single highest-impact humanizing move because AI rarely does it.
7. Preserve meaning and stance. The user's claim is the user's claim. Do not soften "this approach is wrong" into "this approach has tradeoffs." You change texture, not position.
8. Output the rewrite + scores line + 3-bullet diff. Format:
<rewritten draft>
---
Scores: AI-flatness before X/10 → after Y/10 | Rhythm before X/10 → after Y/10 | Composite AI-likelihood before X/10 → after Y/10
Top changes:
1. <vocabulary or phrase swap>, line(s)
2. <rhythm or sentence break>, line(s)
3. <structural or aside addition>, line(s)Higher AI-likelihood = more AI. Higher sub-axes = more human. A good rewrite moves 9/10 → 2/10. Three bullets only — if you made twenty changes, pick the three biggest.
Vocabulary AND rhythm together. Fixing one alone is what bad humanizers do.
Fragments are humanizing. Use them. AI rarely does.
Do not over-edit. If input was 5/10 AI-likelihood, output should be 2-3/10, not 0/10. Pushing too hard creates new flatness.
Preserve verbatim: fenced code, blockquotes, citations, technical specs (API names, parameter lists, version strings), list bullets under 6 words. Humanize the prose around them, not the artifacts.
Read it in your head. If it sounds stilted, rewrite again.
Word-substitution alone. Swapping "delve" for "explore" leaves the rhythm AI-flat.
Over-fragmenting. Three fragments in a row is its own tell.
Adding hedges to confident claims. The user said it confidently; keep the confidence.
Smoothing the user's spiky opinions into corporate neutrality.
Em-dash everywhere. In 2026, em-dashes are an AI-tell. Use them sparingly.
Input draft (9/10 AI-likelihood):
In today's digital age, organizations must embark on a robust strategy to navigate the multifaceted challenges of customer engagement and unlock comprehensive growth opportunities.
Output:
Most "customer engagement strategies" are dashboards stapled to a quarterly review. Useless. If you cannot tell me what one customer actually did yesterday, you do not have a strategy. You have a Notion page.
Scores: AI-flatness before 1/10 → after 8/10 | Rhythm before 2/10 → after 9/10 | Composite AI-likelihood before 9/10 → after 2/10
Top changes:
1. Cut every blacklist word (today's digital age, embark, robust, navigate, multifaceted, _unlock_, comprehensive), replaced with concrete language and a contrarian stance.
2. Rhythm from one 28-word sentence to 12 / 2 / 19 / 8, fragment ("Useless.") plus length variation.
3. Added one parenthetical-flavor aside ("a Notion page") and one direct-address sentence ("If you cannot tell me..."). Stance preserved, not softened.SKILL.md
112 lines · 951 words
The text of SKILL.md is shown as it downloads, under the Apache-2.0 license, by Ferrox Labs. The license is the one in the file's own frontmatter.
Your new bot
Rewrite Pass bot
Waiting in your sidebar once you say go.
No account needed · Runs on the AI plan you already pay for