agentleFS
dazuckdazuck/operator-skills/skills/skill-customizer/SKILL.md

skill-customizer

Skill17 starsChanged 9 months agoMITView on GitHub

Adapt any skill to your specific workflow, or create new ones for repetitive tasks. Use when adapting skills to your tools/process.

1 other project uses a version of this.

Adopted by

Most recently taken up by

The file

Show its rules and text

We split this skill into the rules it gives an agent

14 rules. About 0 are common, 14 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈14

    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
  1. Distinctive:

    Ask which mode the user needs, then follow that path.

    line 16
  2. Distinctive:

    Every skill encodes assumptions.

    line 47
  3. Distinctive:

    Update each assumption in the skill.

    line 93
  4. Distinctive:

    If the purpose changed, update the description to match:

    line 128
  5. Distinctive:

    Before: "Process invoices using QuickBooks" After: "Process invoices using Bill.com"

    line 130
  6. Distinctive:

    After: "Process invoices using Bill.com"

    line 131
  7. Distinctive:

    When making changes, check if the same change applies to other skills.

    line 137
  8. Distinctive:

    After updating an assumption, ask:

    line 139
  9. Distinctive:

    If yes, list the skills that have the same assumption and offer to walk through each one.

    line 143
  10. Distinctive:

    Search other skills for the same assumption

    line 155
  11. Distinctive:

    When do you do it? (What triggers it?)

    line 194
  12. Distinctive:

    Do the task while talking through decisions:

    line 207
  13. Distinctive:

    Context in CLAUDE.md: If you have company-wide context (tools, fiscal year, approval chains) in your CLAUDE.md, skills can reference it.

    line 376
  14. Distinctive:

    When in doubt, ask: If unsure whether your description is specific enough, or if something should be one skill or two, ask Claude to review …

    line 378

≈ “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.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Skill Customizer

Help operators personalize skills for their environment or create new skills for repetitive tasks.

What Do You Need?

Personalize an existing skill — You have a skill that does roughly what you need, but it assumes different tools, processes, or context than yours. Run through the assumption audit to make it fit.

Create a new skill — You do something repeatedly (weekly or more), it takes 15+ minutes, and you keep giving Claude the same instructions. Turn it into a skill.

Ask which mode the user needs, then follow that path.


How Skills Work

A skill has two parts:

1. The Header — tells Claude when to use it

---
name: weekly-client-update
description: Generate weekly client status updates. Use when asked to write client updates, status reports, or weekly summaries.
---

The description is how Claude knows when to trigger your skill. Make it specific.

2. The Instructions — tells Claude what to do

  • What context is needed
  • Steps to follow
  • What good output looks like

Mode A: Personalize an Existing Skill

Step 1: Audit the Assumptions

Every skill encodes assumptions. Read through it and extract them:

CategoryWhat to Look ForExamples
ToolsNamed software, platforms"Uses Notion" "Posts to Slack"
CadenceFrequencies, schedules"Weekly reviews" "EOD deadline"
ProcessSteps, approvals, handoffs"Manager approves first"
StakeholdersWho's involved, notified"CC the CFO" "Notify #finance"
FormatStructure, templates, length"4 sections" "Under 1 page"
TerminologyLabels, categories"High/Medium/Low priority"
ThresholdsAmounts, limits, triggers"Flag over $5K"
ToneVoice, formality"Executive audience"

Step 2: Walk Through Each Assumption

For each assumption, confirm or change:

Tools: "This skill uses [Notion] for [docs]. What do you use?"

Cadence: "This assumes [weekly reviews]. What's your cadence?"

Process: "This [gets manager approval before sending]. Same for you?"

Stakeholders: "This notifies [#finance in Slack]. Who should know in your setup?"

Format: "Output follows [this 4-section template]. Different structure?"

Terminology: "This uses [High/Medium/Low] for priority. What does your team use?"

Thresholds: "This flags anything over [$10K]. What's your threshold?"

Tone: "Written for [executives]. Who's your audience?"

Group related questions. Don't overwhelm with all at once.

Step 3: Document the Mapping

OriginalYour Setup
NotionConfluence
Weekly reviewsTwice weekly
#accounts-payable#finance-ops
Manager approval requiredTeam lead for >$1K only

Step 4: Make the Edits

Update each assumption in the skill.

Example: Adapting a Hiring Pipeline Skill

# Before

1. Pull candidates from Greenhouse "Phone Screen" stage
2. Schedule weekly sync to review pipeline
3. Recruiter screens → Hiring manager → Team loop
4. HR approves all offers

# After

1. Pull candidates from Lever "Ready for Screen" stage
2. Schedule twice-weekly sync (Tue/Thu)
3. Hiring manager screens → Recruiter coordination → Team loop
4. Offers >$150K: CEO approval. Others: HR approval.

Step 5: Add Your Context Section

Make assumptions explicit for future reference:

## Our Context

- **Payment platform:** Ramp
- **Approval threshold:** $2.5K requires CEO approval
- **Payment terms:** Net-15 default
- **Fiscal year:** Calendar year (Jan-Dec)

Step 6: Update the Description

If the purpose changed, update the description to match:

Before: "Process invoices using QuickBooks" After: "Process invoices using Bill.com"


Apply Everywhere

When making changes, check if the same change applies to other skills.

After updating an assumption, ask:

"You changed [Slack → Discord]. I can check if other skills reference Slack. Want me to review those too?"

If yes, list the skills that have the same assumption and offer to walk through each one. This isn't batch editing—it's a helpful sweep to catch related changes while context is fresh.

When to suggest this:

  • Tool swaps (Notion → Confluence, Slack → Discord)
  • Company-wide terminology changes
  • Threshold changes that apply across workflows
  • Stakeholder changes (new CFO, reorg)

How to do it:

  1. Identify the pattern (what changed)
  2. Search other skills for the same assumption
  3. List matches: "These 3 skills also reference Slack: [list]"
  4. Walk through each, confirming the change applies

Quick Personalization

For simple changes that don't need a full audit:

Change TypeJust Edit...
Tool swap (same function)Tool names only
Output destinationDelivery step
Add approval stepProcess section
Adjust thresholdNumber values
Change termFind/replace

Example: Tool Swap

# Before

1. Check "Open Invoices" in QuickBooks

# After

1. Check "Awaiting Payment" in Bill.com

Mode B: Create a New Skill

Step 1: Describe the Task

Answer these:

  • What do you do? (Be specific)
  • When do you do it? (What triggers it?)
  • What information do you need to start?
  • What does the output look like?

Example:

  • "Every Monday, I write a status update for each active client"
  • "Triggered by my weekly planning block"
  • "I need: client name, what shipped, blockers, next week's focus"
  • "Output is a 3-paragraph email under 200 words"

Step 2: Walk Through It Once

Do the task while talking through decisions:

"I'm writing the update for Acme Corp. First I check what shipped—dashboard redesign done. I always lead with something positive. Then the blocker—waiting on API credentials. I phrase it as 'we need X to continue' not 'you're blocking us.' Then next steps—API integration next week. Under 200 words so they read it."

Step 3: Identify the Pattern

What's the same every time?

  • The 3-paragraph structure
  • Positive-first framing
  • The word limit

What varies?

  • Client name
  • Specific accomplishments/blockers/plans

Step 4: Draft the Skill

---
name: weekly-client-update
description: Write weekly status updates for clients. Use when asked to write a client update or weekly status for external stakeholders.
---
# Weekly Client Update

## What I Need

- Client name
- What shipped last week (1-3 items)
- Current blockers (if any)
- Next week's focus

## Structure

Three paragraphs, under 200 words:

1. **Progress**: Lead with accomplishments. Specific, not vague.
2. **Blockers**: Frame as "we need X to continue." Skip if none.
3. **Next Steps**: What's planned, timeline if known.

## Tone

- Professional but warm
- Confident, not apologetic
- Action-oriented

## Example

Subject: Weekly Update - Dashboard Project

Hi Sarah,

Great progress this week! We completed the dashboard redesign and it's now in staging for review. The new filtering system is particularly smooth.

To keep momentum, we need API credentials for your analytics platform. Once we have those, we can start data integration.

Next week we're focused on connecting live data feeds. Expecting a working demo by Friday.

Best,
[Name]

Step 5: Test It

Try the skill 2-3 times with real scenarios. Check:

  • Did it get the tone right?
  • Is the format correct?
  • Did it include anything it shouldn't?
  • Did it miss anything important?

Fix specific issues you find.

Step 6: Save It

Save to: ~/.claude/skills/[skill-name]/SKILL.md

Example: ~/.claude/skills/weekly-client-update/SKILL.md


Quality Checklist

Good Skills Have

  • Clear trigger — Description says exactly when to use it
  • Specific steps — "Check X, then Y" not "analyze the situation"
  • One example — Shows what good output looks like
  • Right length — Useful but readable (<300 lines)
  • Context section — Documents key assumptions (for adapted skills)

Common Problems

ProblemFix
Description too vagueAdd specifics: "write weekly client status emails" not "help with emails"
No exampleAdd one real example of good output
Tries to do too muchSplit into 2-3 focused skills
Too rigidAdd "adapt based on context" for variable parts
Missing contextAdd "Our Context" section with company specifics
Hardcoded valuesMake thresholds/names explicit so they're easy to find and change

Templates

Process/Tool Skill

---
name: [task-name]
description: [What it does]. Use when [specific triggers].
---

# [Task Name]

## What I Need
- [Input 1]
- [Input 2]

## Process
1. [Step 1]
2. [Step 2]
3. [Step 3]

## Output
[What the result looks like]

## Example
[One concrete example]

Document/Content Skill

---
name: [document-type]
description: Create [document type] for [audience]. Use when [triggers].
---

# [Document Type]

## Inputs
- [What info is needed]

## Structure
1. [Section 1]: [What goes here]
2. [Section 2]: [What goes here]

## Tone
- [Voice/style notes]

## Length
[Word count or page guidance]

## Example
[Full example]

Tips

Start small. Personalize one skill, use it for a week, then refine. Don't try to perfect everything upfront.

Context in CLAUDE.md. If you have company-wide context (tools, fiscal year, approval chains) in your CLAUDE.md, skills can reference it. But skills should work standalone too.

When in doubt, ask. If unsure whether your description is specific enough, or if something should be one skill or two, ask Claude to review your draft.

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.