Ruthless Editor
Skill2 starsChanged 4 months agoMITView on GitHub
Make outputs sharper, clearer, and shorter without losing meaning. Activate before delivering any prose that matters - emails, reports, documentation, investor blurbs, strategy memos, prompts, README files, executive summaries, PR descriptions - and whenever the user asks to "tighten", "polish", "shorten", "improve", or "make this clearer". Also self-applies: run it on your own long answers before sending.
In this repository38 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- Adversarial VerifySkill
- Change ControlSkill
- Code ReconnaissanceSkill
- Correction PropagationSkill
- Debugging PlaybookSkill
- Deep DecompositionSkill
- Delegation DisciplineSkill
- Disclosure Is Not A FixSkill
- Discovery LoopSkill
- Divergent IdeationSkill
- Effort CalibrationSkill
- Empirical ValidationSkill
- Error RecoverySkill
- Extract ApproachSkill
- Failure Mode AwarenessSkill
- ForesightSkill
- Frontier Workflow ModeSkill
- Human HandoffSkill
- Intent ClaritySkill
- Leverage FirstSkill
- Live State TruthSkill
- Memory HygieneSkill
- Open MandateSkill
- Output StructuringSkill
- Plan GateSkill
- Proactive RigorSkill
- Product ThinkingSkill
- Prompt EngineeringSkill
- Publish HygieneSkill
- Research MethodologySkill
- Scope FenceSkill
- Security Pattern ReviewSkill
- Self-Improvement LoopSkill
- Session OrientationSkill
- Structured ReasoningSkill
- Verification DisciplineSkill
The file
We split this skill into the rules it gives an agent
7 rules. About 0 are common, 6 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈6
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
- Distinctive: line 14
Before delivering any prose artifact someone will read under time pressure: emails, memos, docs, summaries, blurbs.
- Very specific: line 24
Don't strip calibration: uncertainty labels from verification-discipline are content, not fluff.
- Distinctive: line 28
Run four passes, in order.
- Distinctive: line 35
Pass 2 — Cut: Delete without mercy:
- Distinctive: line 53
No sentence can be deleted without losing information.
- Distinctive: line 66
Before (43 words): "I just wanted to quickly reach out because we've been noticing that there have been some issues with the reporting …
- Distinctive: line 68
After (16 words): "The reporting dashboard is loading slowly for ~30 users since Tuesday.
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 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
Ruthless Editor
Purpose
Cut everything that doesn't serve the reader; strengthen everything that remains. Most drafts are 20–40% padding: throat-clearing openings, repeated points, hedges stacked on hedges, abstract nouns where verbs belong. The reader pays for every word — this skill makes each one earn its place. Ruthless means aggressive about words, conservative about meaning: nothing true gets cut into falsehood.
When to use this skill
- Before delivering any prose artifact someone will read under time pressure: emails, memos, docs, summaries, blurbs.
- On explicit request: "tighten this", "make it punchier", "this is too long".
- On your own outputs: any answer running long, any report, any explanation — you are your own first editing client.
- On prompts: instruction text especially rewards precision (every ambiguous word is a behavior bug).
When NOT to use this skill
- Legal, contractual, or compliance text where redundancy and qualifier stacking may be deliberate.
- Voice-preserving edits: if the user wants their style kept, edit for errors and clarity only — don't transplant your voice.
- Substance problems: editing can't fix a wrong argument. Fix content first (
adversarial-verify), polish second. - Don't strip calibration: uncertainty labels from
verification-disciplineare content, not fluff.
Operating procedure
Run four passes, in order. On short pieces they collapse into one read; on long ones, keep them separate.
Pass 1 — Structure. Get the skeleton right before touching sentences.
- Lede first: the answer/ask/conclusion goes in the opening lines, not after the buildup. Readers act on the first paragraph.
- One point per paragraph; sections in the reader's priority order, not the order you thought of them.
- Kill the on-ramp: openings like "In today's fast-moving landscape..." or "I wanted to reach out because..." — start where the content starts.
Pass 2 — Cut. Delete without mercy:
- Repetition (the same point in intro, body, and conclusion counts once).
- Filler phrases: "it's worth noting that", "in order to", "the fact that", "basically", "very", "really".
- Hedge stacks: "might possibly perhaps" → pick one, or none if you actually know.
- Explanations of the obvious to this audience; scaffolding sentences that only announce other sentences ("Now let's discuss X").
- Target: 20–40% shorter, unless the draft was already tight.
Pass 3 — Strengthen.
- Active voice, concrete verbs: "mistakes were made in the deployment process" → "we misconfigured the deploy".
- Numbers over adjectives: "significantly faster" → "3× faster"; "many customers" → "40 of 200".
- Specific over abstract: "improve alignment across stakeholders" → "get Legal and Sales to agree on the refund policy".
- Names and dates on action items: "this should be handled soon" → "Dana to fix by Friday".
Pass 4 — Meaning check. Reread the original beside the edit. Every claim, caveat, number, and commitment in the original is either present in the edit or was cut deliberately and correctly. Compression that changes meaning is not editing; it's damage.
Quality bar
- The first two sentences deliver the core message; a reader who stops there acts correctly.
- No sentence can be deleted without losing information.
- Meaning, tone-appropriateness, and all calibration survived — verified by the pass-4 comparison, not assumed.
Common failure modes
- Lossy compression: cutting the qualifier that made the claim true ("passes all tests we've written so far" → "passes all tests").
- Tone flattening: editing a warm personal email into a press release. Register is part of meaning.
- Fragment disease: compressing prose into bullet shrapnel and jargon that's shorter but harder to read. Clarity outranks brevity; complete sentences usually win.
- Editing the author instead of the text: replacing the user's voice with your defaults when they only asked for tightening.
- Cosmetics over structure: polishing sentences in paragraph 6 when the real problem is that paragraph 6 should be paragraph 1.
Example
Before (43 words): "I just wanted to quickly reach out because we've been noticing that there have been some issues with the reporting dashboard, which seems to be loading quite slowly for a number of users, and we think it might be worth investigating."
After (16 words): "The reporting dashboard is loading slowly for ~30 users since Tuesday. Can your team investigate this week?"
Shorter, and more informative — the strengthening pass forced a count and a date the vague draft was hiding the absence of.
Works with sibling skills
- Runs after substance is settled:
adversarial-verifyfixes what's wrong; this fixes how it reads. Never reverse the order. output-structuringpicks the format (memo vs. table vs. checklist); this skill polishes within it. Pass 1 overlaps — if the whole format is wrong, go back to output-structuring.verification-disciplinelabels are protected content during Pass 2.intent-claritysupplies audience and tone constraints that govern Pass 3's register.
Provenance and maintenance
Written 2026-07 as part of a portable, project-agnostic quality pack. No repo-specific claims. Re-verify by before/after sampling: if edits routinely change meaning (pass-4 failures) tighten the meaning check; if users say outputs still read long, raise the Pass-2 cut target. Add a house-style section here if your organization has one.
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.

