Learn from Conversation

Analyze the conversation to extract lessons learned, then persist them to AI assistant configuration files.

Arguments

Optional text to narrow what to learn (e.g. sandbox workaround, that build fix).

If $ARGUMENTS is help, --help, -h, or ?, skip the workflow and read references/options.md.

Supported Assistants

Assistant	Config File	Format
Claude Code	CLAUDE.md	Markdown
Gemini	GEMINI.md	Markdown
AGENTS.md	AGENTS.md	Markdown
Cursor	.cursorrules or .cursor/rules/*.mdc	Markdown/MDC
GitHub Copilot	.github/copilot-instructions.md	Markdown
Windsurf	.windsurf/rules/rules.md	Markdown
Continue	.continuerc.json	JSON

See references/assistant-configs.md for format details.

Process

1. Detect and Assess Configurations

Scan for config files and check their sizes:

for f in CLAUDE.md GEMINI.md AGENTS.md .cursorrules .github/copilot-instructions.md \
  .windsurf/rules/rules.md .continuerc.json; do
  [ -f "$f" ] && wc -l "$f"
done
find .cursor/rules -name "*.mdc" -exec wc -l {} \; 2>/dev/null

Behavior based on detection:

Scenario	Action
Single config found	Update it automatically
Multiple configs found	Prompt user to select which to update
No configs found	Display init commands from assistant-configs.md, then exit

Size Thresholds

Lines	Status	Action
< 400	Healthy	Add learnings directly
400-500	Warning	Add carefully, suggest cleanup
> 500	Oversized	Refactor before adding new content

Use these same thresholds throughout (routing decisions, present-and-confirm, edge cases).

Discover Existing Skills

List skills for routing decisions:

find . -name "SKILL.md" -type f 2>/dev/null | grep -v node_modules | \
  xargs grep -l "^name:" | while read -r f; do
  grep -m1 "^name:" "$f" | sed 's/name: //'
done

2. Analyze Conversation

Scan for:

• Corrections: Commands retried, assumptions proven wrong, missing prerequisites
• Discoveries: Undocumented patterns, integration quirks, environment requirements
• Improvements: Steps that should be automated or validated earlier
• Instruction Violations: Existing config file guidelines that were not followed during the conversation. If the user had to remind the assistant to do something already documented, flag which section was ignored and assess whether the wording needs to be more prominent or actionable

3. Categorize and Route Each Learning

Category	Primary Destination	Fallback When Oversized
Project facts	Config file	Extract to new skill
Prerequisites	Config file	Extract to project-setup skill
Environment	Config file	Extract to environment-setup skill
Workflow pattern	Existing related skill	Create new skill
Automated workflow	New skill	New skill (always)

Routing Decision Tree

For each learning, evaluate in order:

1. Is this a multi-step automated workflow (>5 steps)?

   • YES → Create new skill (proceed to Step 5, Route C)
   • NO → Continue

2. Does an existing skill cover this topic?

   • YES → Update that skill
   • NO → Continue

3. Is the target config file oversized (>500 lines)?

   • YES → Create new skill OR offer refactoring (see Step 4)
   • NO → Continue

4. Is this learning situation-specific (applies to narrow context)?

   • YES → Create new skill with globs or context constraints
   • NO → Add to config file

Size-Based Rules

Learning Size	Preferred Destination
< 3 lines	Config file (even if near threshold)
3-30 lines	Follow decision tree above
> 30 lines	Strongly prefer skill creation

4. Present and Confirm

For each learning, show:

**[Category]**: [Brief description]
- Source: [What happened in conversation]
- Proposed change: [Exact text or file to add]
- Destination: [Config file] ([current] → [projected] lines)

Handle Size Threshold

If adding the learning would push a config file over threshold:

Adding this learning would bring [filename] to [X] lines (threshold: [Y]).

Options:
1. Add learning anyway (not recommended)
2. [Refactor](references/refactoring.md) existing content to skills first, then add
3. Create a new skill for this learning instead
4. Skip this config file

Ask for confirmation before applying each change.

5. Apply Changes

Apply changes based on routing decision from Step 3:

Route A: Add to Config File

Markdown configs (CLAUDE.md, GEMINI.md, AGENTS.md, Copilot, Windsurf):

• Find appropriate section, preserve existing structure
• Append to relevant section or create new section if needed

Cursor rules:

• Legacy .cursorrules: Treat like markdown, append content
• Modern .cursor/rules/*.mdc: See references/format-cursor-mdc.md

Continue (.continuerc.json):

• Update customInstructions field, preserving existing content
• See references/format-continue.md

Route B: Update Existing Skill

When adding to an existing skill:

1. Read the skill file: skills/[name]/SKILL.md
2. Find appropriate section or create new one
3. Append learning, maintaining the skill's existing structure
4. If skill has references, consider adding to reference file instead

Skill update format:

## [New Section or append to existing]

[Learning content formatted as guidance or workflow step]

Route C: Create New Skill

Create in skills/[name]/SKILL.md with this template:

---
name: [learning-topic]
description: [What this handles and when to use it - triggers belong here, not in body]
---

# [Learning Topic]

[Learning content structured as workflow]

## Process

### 1. [First Step]
[Details]

6. Summarize

List:

• Config files modified (with full paths)
• Sections updated in each file
• Any skills created

Examples

Situation	Learning
"e2e tests failed because API wasn't running"	Add prerequisite to selected config(s)
"Parse SDK doesn't work with Vite out of the box"	Document workaround in selected config(s)
"Build failed because NODE_ENV wasn't set"	Add required env var to selected config(s)
"Every component needs tests, lint, build..."	Create add-component skill
"User asked to update README but config already requires it"	Flag instruction violation — fix wording or prominence, not add new rule

Edge Cases

Scenario	Handling
No configs detected	Guide user to initialize one first, exit early
Multiple configs found	Prompt user to select which to update
Malformed config file	Warn and skip that file
Duplicate content exists	Check before adding, warn if similar learning exists
Config file already oversized	Offer refactoring before adding (Step 4)
Learning matches multiple skills	Present options, let user choose which skill to update
Skill file also oversized	Suggest creating sub-skills or reference files
No existing skills found	Skip skill matching, proceed with config or new skill

Guidelines

• Be minimal: Only add what genuinely helps future sessions
• Avoid duplication: Check for existing similar content before adding
• Prefer specificity: "Run npm run dev before e2e tests" beats "ensure services are running"
• Focus on non-obvious: Skip things Claude would naturally do; capture what surprised you