writing-principles
Skill0 starsChanged 6 months agoMITView on GitHub
Enforces writing principles for instruction documents (SKILL.md, skill supporting files, subagents, CLAUDE.md). Use when writing, reviewing, or editing any file in **/skills/**/*.md, **/agents/*.md, or **/CLAUDE.md. Triggers on skill creation, skill file updates, CLAUDE.md changes, subagent authoring, or instruction document review.
We split this skill into the rules it gives an agent
7 rules. About 0 are common, 7 are distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈7
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 24
Gate: Do not claim work complete until self-check passes run at calibrated rigor level
- Distinctive: line 26
Gate: Do not fix violations until user provides direction
- Distinctive: line 74
After writing or editing, run passes appropriate to risk level.
- Distinctive: line 93
High risk: All passes
- Distinctive: line 99
After completing the self-check:
- Distinctive: line 103
Ask for guidance before making fixes
- Distinctive: line 105
Do not auto-fix. Wait for user direction.
≈ “Common” is an estimate: at least 1 in 5 of the 308,232 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
Writing Principles
Enforce consistent quality in instruction documents by applying the 14 writing principles.
Scope
This skill applies when you write, review, or edit:
**/skills/**/*.md— Skill files and supporting documentation**/agents/*.md— Subagent instruction files**/CLAUDE.md— Project and user instruction files
Workflow
- Recognize — Identify you're working on an instruction document
- Calibrate — Assess document risk level (Low/Medium/High)
- Apply — Use principles while writing or editing
- Verify — Run self-check passes appropriate to risk level
- Gate: Do not claim work complete until self-check passes run at calibrated rigor level
- Report — List violations found and ask for guidance before fixing
- Gate: Do not fix violations until user provides direction
Risk Calibration
Assess risk level before writing. Risk level = highest factor.
| Factor | Low | Medium | High |
| Scope | Personal preferences | Project defaults | Multi-agent workflows |
| Reversibility | Easy to change | Requires coordination | Affects downstream systems |
| Ambiguity tolerance | High (preferences) | Medium (conventions) | Low (procedures) |
| Typical length | <50 lines | 50-150 lines | >150 lines |
If uncertain: treat as Medium.
Rigor by Risk Level
| Risk | Self-Check Passes | Iteration Limit |
| Low | Passes 1-3 (items 1-15) | 1 |
| Medium | Passes 1-7 (items 1-27) | 2 |
| High | All passes (items 1-52) | 5 |
Principles Quick Reference
Apply while writing. Lower number = higher priority in conflicts.
| # | Principle | Core Rule | Red Flag |
| 1 | Be Specific | Replace vague language with concrete values | Vague pronouns, hedge words, unspecified quantities |
| 2 | Define Terms | Explain jargon and acronyms on first use | Unexplained acronyms, assumed project knowledge |
| 3 | Show Examples | Illustrate rules with concrete instances | Rules without demonstration, abstract patterns |
| 4 | Verify Interpretation | Include confirmation checkpoints for high-risk instructions | No verification for ambiguous scope, irreversible actions |
| 5 | State Boundaries | Explicitly declare scope and mutability | Implicit "obvious" scope, unstated read-only |
| 6 | Specify Failure Modes | Define behavior when preconditions fail | Happy-path-only instructions, vague error handling |
| 7 | Specify Defaults | State behavior when no instruction applies | Implicit defaults, unhandled case improvisation |
| 8 | Declare Preconditions | State requirements and verification before execution | Assumed working directory, tools, or state |
| 9 | Close Loopholes | Anticipate and block creative misinterpretations | Rules without rationale, unaddressed edge cases |
| 10 | Front-Load | Put critical information first | Commands buried after context |
| 11 | Group Related | Keep conditions near consequences | Cross-references, scattered related content |
| 12 | Keep Parallel | Match structure across similar content | Mixed voice in lists, inconsistent hierarchy |
| 13 | Specify Outcomes | Define observable success criteria | "Ensure it works," process without verification |
| 14 | Economy | Remove words that don't advance meaning; use active voice | Filler phrases, passive voice, double negatives |
For full principle details, examples, and self-check procedure: writing-principles.md
Self-Check Procedure
After writing or editing, run passes appropriate to risk level.
Pass Overview
| Pass | Focus | Items | Priority |
| 1 | Specificity | 1-4 | 1 |
| 2 | Terms and Examples | 5-7 | 1 |
| 3 | Verification and Authority | 8-15 | 1 |
| 4 | Boundaries | 16-18 | 2 |
| 5 | Preconditions, Failure, Defaults | 19-31 | 2 |
| 6 | Loopholes | 32-34 | 3 |
| 7 | Structure and Front-Loading | 35-40 | 4 |
| 8 | Outcomes | 41-43 | 5 |
| 9 | Economy | 44-49 | 6 |
| 10 | Coherence | 50-53 | — |
Low risk: Passes 1-3 only (Priority 1) Medium risk: Passes 1-6 (Priority 1-3) High risk: All passes
For the full 53-item checklist: writing-principles.md
Reporting Violations
After completing the self-check:
- List each violation with principle name and location
- Format:
"[Principle #X]: [description] at [location]" - Ask for guidance before making fixes
Do not auto-fix. Wait for user direction.
Red Flags
These thoughts mean STOP — you're rationalizing skipping the principles:
| Thought | Reality |
| "This is a small edit" | Small edits compound. Apply principles. |
| "The document is already good" | Run the self-check anyway. |
| "I know these principles" | Knowing ≠ applying. Run the check. |
| "This will slow things down" | Violations cause rework. Take the time. |
| "The user is in a hurry" | Fast and wrong wastes more time. |
| "It's just a supporting file" | Supporting files are instruction documents too. |
Failure Mode Index
Maps failure modes to preventing principles. Use for gap analysis.
| Failure Mode | Preventing Principles | Notes |
| Ambiguity causing wrong behavior | #1 (Specific), #2 (Terms), #3 (Examples) | Priority 1 cluster |
| Misinterpretation of high-risk actions | #4 (Verify Interpretation) | Checkpoint pattern |
| Scope creep / boundary violation | #5 (Boundaries) | Inclusion + exclusion |
| Improvised error handling | #6 (Failure Modes) | Explicit failure paths |
| Undefined default behavior | #7 (Defaults) | Global fallback |
| Runtime precondition discovery | #8 (Preconditions) | Requires/Check pattern |
| Version drift / resource staleness | #8 (Preconditions) | Version and Freshness subsection |
| Creative non-compliance | #9 (Loopholes) | Name and close rationalizations |
| Deprioritized information | #10 (Front-Load) | Lead with conclusions |
| Scattered reassembly | #11 (Group Related) | Physical = logical distance |
| Parsing overhead | #12 (Parallel) | Consistent structure |
| Unverifiable success | #13 (Outcomes) | Observable criteria |
| Signal dilution / context waste | #14 (Economy) | Remove non-advancing words |
Other versions3 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 3
- haowjy/creative-writing-skills★ 555same name
- FiredMosquito831/my-writing-companion★ 2same name
- Abdullokh-N/proza-ru★ 0same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository5 other files it gives its agents.
- git-hygieneSkill
- making-recommendationsSkill
- next-stepsSkill
- scrutinizeSkill
- system-design-reviewSkill
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
Reviews can't be read right now.
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.