prompt-rewriter
Skill0 starsChanged 7 months agoMITView on GitHub
Rewrite system prompts to strip compensating complexity and produce clean outcome-based prompts. Takes audit output (from compensating-complexity-auditor or manual analysis) and produces a four-component prompt — outcome spec, constraints, tools, coordination pattern. Use when modernizing prompts after a model upgrade, simplifying over-engineered system prompts, or converting procedural instructions to outcome-based ones. Trigger on "rewrite prompt", "simplify prompt", "strip scaffolding", "outcome-based prompt", "modernize prompt".
1 other project uses a version of this.
Adopted by
Most recently taken up by
- novahiz/novahiztheir own version, first seen Oct 11, 2026
Other versions1 project uses a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 1
- novahiz/novahiz★ 1same 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 repository24 other files it gives its agents.
- aarSkill
- agent-cost-modelSkill
- banana-makerSkill
- bitter-lesson-scorecardSkill
- boot-tax-monitorSkill
- compensating-complexity-auditorSkill
- context-fork-guideSkill
- context-hygieneSkill
- counterargument-stress-testSkill
- dark-code-auditSkill
- executive-briefingSkill
- failure-asymmetrySkill
- failure-postmortemSkill
- file-intelSkill
- geopolitical-signal-enricherSkill
- get-api-docsSkill
- gh-reviewSkill
- model-routerSkill
- optimize-descriptionSkill
- skill-auditSkill
- skill-maintenanceSkill
- spec-gap-detectorSkill
- token-burn-auditorSkill
- what-am-i-forgettingSkill
The file
We split this skill into the rules it gives an agent
5 rules. About 0 are common, 4 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈4
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 17
If no audit exists, do a fast classification pass (outcome/constraint/scaffolding/duct-tape) before rewriting.
- Distinctive: line 52
No imperative procedures ("First do X, then do Y")
- Distinctive: line 71
All outcome logic from the original must be preserved
- Distinctive: line 72
All hard constraints must be preserved
- Distinctive: line 73
No procedural scaffolding in the output
≈ “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
Outcome-Based Prompt Rewriter
Transform scaffolding-heavy system prompts into clean, outcome-based prompts that work with capable models.
Phase 1: Input
Accept ONE of:
- Raw audit output from the
compensating-complexity-auditorskill (preferred) - A system prompt the user wants rewritten (will do a quick classification pass first)
- A file path to a system prompt or CLAUDE.md
If no audit exists, do a fast classification pass (outcome/constraint/scaffolding/duct-tape) before rewriting.
Phase 2: Extract Essential Intent
From the classified components, extract:
- Outcomes — What must the system achieve? (strip all "how")
- Hard constraints — Business rules, safety, compliance, quality thresholds
- Available tools — What tools/APIs/resources does the system have access to?
- Coordination pattern — If multi-agent: who talks to whom, what are handoff conditions?
Discard:
- Step-by-step procedures (let the model figure out "how")
- Format micromanagement (unless format IS the business requirement)
- Retry/fallback logic for known model weaknesses
- Few-shot examples that teach capabilities the model already has
Phase 3: Rewrite
Produce a clean prompt with exactly four sections:
## Outcome
[What this system must achieve — measurable, specific]
## Constraints
[Hard rules that cannot be violated — business, safety, quality]
## Tools
[Available tools and when to use them — not how to use them]
## Coordination
[Multi-agent handoffs, escalation conditions — if applicable]Rules for the rewrite:
- No imperative procedures ("First do X, then do Y")
- No "you are a..." role-play preambles unless they genuinely change behavior
- Constraints only include things that are NOT obvious to a capable model
- Tool descriptions say WHAT the tool does, not step-by-step usage
- Total length should be 30-60% of the original (if it's longer, you added complexity)
Phase 4: Diff + Risk Assessment
Show a before/after comparison:
- What was kept (and why)
- What was deleted (and why)
- What was reworded (and what changed)
- Risk items: anything deleted that MIGHT still be needed — flag these as "test before deploying"
Output a length comparison: original line count vs. rewritten line count, with percentage reduction.
Verification
- Rewritten prompt must be shorter than the original
- All outcome logic from the original must be preserved
- All hard constraints must be preserved
- No procedural scaffolding in the output
- Risk items explicitly flagged for testing
Source
Nate's Newsletter (2026-04-01): Outcome-based prompt rewriting as the second step of the compensating complexity removal pipeline. Pairs with compensating-complexity-auditor.
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.

