chat-to-skill
Skill57 starsChanged 5 months agoMITView on GitHub
Convert current chat session into a reusable skill. Use when user says "/chat-to-skill", "save this as skill", "create skill from chat", "turn this into a skill", or wants to preserve learnings from the conversation as long-term memory.
In this repository15 other files it gives its agents.
- AGENTS.mdAGENTS.md
- agent-browserSkill
- ai-newsSkill
- applescriptSkill
- birdSkill
- dont-anger-marioSkill
- github-issuesSkill
- keyboard-layout-decoderSkill
- maintainer-replySkill
- package-releaseSkill
- pr-reviewSkill
- session-reflectSkill
- skill-discoverySkill
- twitter-thread-styleSkill
- vibe-mergeSkill
The file
We split this skill into the rules it gives an agent
9 rules. About 0 are common, 8 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈8
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 42
Remove project names, organization names, specific entities
- Distinctive: line 44
If something is just an implementation detail (name parsing, date formatting), it's not a separate skill
- Distinctive: line 63
Search strategies that worked
- Distinctive: line 80
If dialog contains several distinct learnable patterns, use AskUserQuestion:
- Distinctive: line 95
If NO to any — either abstract further or skip skill creation.
- Distinctive: line 140
If rarely — maybe no skill needed.
- Distinctive: line 146
Trigger: When should this skill activate?
- Distinctive: line 157
No exploration, only essentials
- Very specific: line 168
Use /skill-creator with the prepared knowledge package.
≈ “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
Chat to Skill
Transform conversation history into reusable skills — long-term memory for Claude.
Process
1. Analyze Dialog
Scan the entire conversation to identify:
- Primary goal: What was the user trying to achieve?
- Secondary goals: Any related objectives discovered along the way
- Errors encountered: Mistakes, dead ends, wrong approaches
- Successful path: What actually worked
2. Abstract to Reusable Patterns
Critical: Do NOT create skills for specific cases. Abstract to general patterns.
Ask yourself:
- What CATEGORY of problem was solved? (not the specific instance)
- What would this look like with different data/context?
- Would this skill be useful in other projects?
Abstraction levels (from bad to good):
| Too specific (BAD) | Good abstraction |
| "Seed users from client Excel" | "Import spreadsheet data into Rails" |
| "Parse names into fields" | (implementation detail, not a skill) |
| "Fix pytest in project X" | "Configure pytest for monorepos" |
| "Add dark mode to app Y" | "Implement theme switching in React" |
Rules:
- Remove project names, organization names, specific entities
- Focus on the TECHNIQUE, not the specific data
- If something is just an implementation detail (name parsing, date formatting), it's not a separate skill
- One dialog = usually one skill (the main workflow), not multiple micro-skills
3. Extract Context-Specific Details
Depending on the task type, look for:
Development tasks:
- Commands and flags that worked
- Versions and compatibility (what works with what)
- Configuration that was needed
- Code patterns and architectural decisions
- Debugging process (how the root cause was found)
- Tool/library choices and why
Research/analysis tasks:
- Sources that proved useful
- Search strategies that worked
- How to validate findings
Process/workflow tasks:
- Order of operations (what must come first)
- Decision criteria (how choices were made)
- Stakeholders or dependencies
Any task:
- Prerequisites that weren't obvious
- Context that matters for success
- Signs that indicate the right/wrong path
4. Choose Topic (if multiple found)
If dialog contains several distinct learnable patterns, use AskUserQuestion:
Question: "I found several skill candidates. Which to create?"
Options: List 2-4 abstracted topics (not specific tasks), plus "All of them"
5. Validate Skill Candidate
Before proposing, check:
- Is this reusable in other projects/contexts?
- Is this a workflow/technique, not just a one-off fix?
- Would future-me benefit from having this skill?
- Is it abstracted enough to apply broadly?
If NO to any — either abstract further or skip skill creation.
6. Propose Topic (if valid)
Present ONE main skill (not a list of micro-topics):
Skill candidate: [Abstracted name]
What it captures: [1-2 sentences about the reusable pattern]
Key learnings:
- [Main insight 1]
- [Main insight 2]
- [Mistake to avoid]Then ask using AskUserQuestion tool:
Question: "How does this skill proposal look?"
Options:
- "Good, create it" — proceed to step 7
- "Too specific" — re-abstract: remove project/entity names, find broader pattern
- "Wrong focus" — re-analyze: what was the MAIN technique vs implementation details?
- "Not reusable" — reconsider: is this a one-off task or a recurring pattern?
7. Revise (if needed)
Based on user's choice, fix the specific issue:
"Too specific":
- What broader category does this belong to?
- Would this apply with different tools/platforms? Generalize.
"Wrong focus":
- What would you google to solve this problem?
- The answer is probably the real skill name.
"Not reusable":
- How often would this exact situation repeat?
- If rarely — maybe no skill needed. Ask user what they hoped to capture.
8. Prepare Knowledge Package
Goal: One sentence — what this skill helps achieve
Trigger: When should this skill activate? (keywords, file types, contexts)
Mistakes to avoid:
- Specific errors from the dialog
- Why they were wrong
- What they cost (time, confusion)
Optimal path:
- Shortest working sequence
- No exploration, only essentials
- Include specific commands/configs if applicable
Key details:
- Versions/compatibility if relevant
- Non-obvious prerequisites
- How to verify success
9. Create Skill
Use /skill-creator with the prepared knowledge package.
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.

