promote – Agents & Automation skill for AI agents
Graduate a proven pattern from auto-memory (MEMORY.md) to CLAUDE.md or .claude/rules/ for permanent enforcement. Use when the user runs /si:promote or asks to make a learned behavior permanent.
Install the promote skill with one command
Pick the tool you use, copy the command and paste it in your terminal. The skill lands in that tool's skills folder and is ready in your next session.
Claude Code
mkdir -p ~/.claude/skills && curl -fsSL https://skill.dn1nguoi.com/z/promote-afb02.tgz | tar xz -C ~/.claude/skills
Call it in Claude Code: /promote – then describe the task you want done.
Codex
mkdir -p ~/.codex/skills && curl -fsSL https://skill.dn1nguoi.com/z/promote-afb02.tgz | tar xz -C ~/.codex/skills
Call it in Codex: $promote – then describe the task you want done.
Hermes Agent
mkdir -p ~/.hermes/skills && curl -fsSL https://skill.dn1nguoi.com/z/promote-afb02.tgz | tar xz -C ~/.hermes/skills
Call it in Hermes Agent: /promote – then describe the task you want done.
DeepSeek
mkdir -p ~/.deepcode/skills && curl -fsSL https://skill.dn1nguoi.com/z/promote-afb02.tgz | tar xz -C ~/.deepcode/skills
Call it in DeepSeek: /promote – then describe the task you want done.
SKILL.md content
/si:promote — Graduate Learnings to Rules
Moves a proven pattern from Claude's auto-memory into the project's rule system, where it becomes an enforced instruction rather than a background note.
Usage
/si:promote <pattern description> # Auto-detect best target
/si:promote <pattern> --target claude.md # Promote to CLAUDE.md
/si:promote <pattern> --target rules/testing.md # Promote to scoped rule
/si:promote <pattern> --target rules/api.md --paths "src/api/**/*.ts" # Scoped with pathsWorkflow
Step 1: Understand the pattern
Parse the user's description. If vague, ask one clarifying question:
- "What specific behavior should Claude follow?"
- "Does this apply to all files or specific paths?"
Step 2: Find the pattern in auto-memory
# Search MEMORY.md for related entries
MEMORY_DIR="$HOME/.claude/projects/$(pwd | sed 's|/|%2F|g; s|%2F|/|; s|^/||')/memory"
grep -ni "<keywords>" "$MEMORY_DIR/MEMORY.md"Show the matching entries and confirm they're what the user means.
Step 3: Determine the right target
| Pattern scope | Target | Example |
|---|---|---|
| Applies to entire project | ./CLAUDE.md | "Use pnpm, not npm" |
| Applies to specific file types | .claude/rules/<topic>.md | "API handlers need validation" |
| Applies to all your projects | ~/.claude/CLAUDE.md | "Prefer explicit error handling" |
If the user didn't specify a target, recommend one based on scope.
Step 4: Distill into a concise rule
Transform the learning from auto-memory's note format into CLAUDE.md's instruction format:
Before (MEMORY.md — descriptive):
> The project uses pnpm workspaces. When I tried npm install it failed. The lock file is pnpm-lock.yaml. Must use pnpm install for dependencies.
After (CLAUDE.md — prescriptive):
## Build & Dependencies
- Package manager: pnpm (not npm). Use `pnpm install`.Rules for distillation:
- One line per rule when possible
- Imperative voice ("Use X", "Always Y", "Never Z")
- Include the command or example, not just the concept
- No backstory — just the instruction
Step 5: Write to target
For CLAUDE.md:
1. Read existing CLAUDE.md
2. Find the appropriate section (or create one)
3. Append the new rule under the right heading
4. If file would exceed 200 lines, suggest using .claude/rules/ instead
For .claude/rules/:
1. Create the file if it doesn't exist
2. Add YAML frontmatter with paths if scoped
3. Write the rule content
---
paths:
- "src/api/**/*.ts"
- "tests/api/**/*"
---
# API Development Rules
- All endpoints must validate input with Zod schemas
- Use `ApiError` class for error responses (not raw Error)
- Include OpenAPI JSDoc comments on handler functionsStep 6: Clean up auto-memory
After promoting, remove or mark the original entry in MEMORY.md:
# Show what will be removed
grep -n "<pattern>" "$MEMORY_DIR/MEMORY.md"Ask the user to confirm removal. Then edit MEMORY.md to remove the promoted entry. This frees space for new learnings.
Step 7: Confirm
✅ Promoted to {{target}}
Rule: "{{distilled rule}}"
Source: MEMORY.md line {{n}} (removed)
MEMORY.md: {{lines}}/200 lines remaining
The pattern is now an enforced instruction. Claude will follow it in all future sessions.Promotion Decision Guide
Promote when:
- Pattern appeared 3+ times in auto-memory
- You corrected Claude about it more than once
- It's a project convention that any contributor should know
- It prevents a recurring mistake
Don't promote when:
- It's a one-time debugging note (leave in auto-memory)
- It's session-specific context (session memory handles this)
- It might change soon (e.g., during a migration)
- It's already covered by existing rules
CLAUDE.md vs .claude/rules/
| Use CLAUDE.md for | Use .claude/rules/ for |
|---|---|
| Global project rules | File-type-specific patterns |
| Build commands | Testing conventions |
| Architecture decisions | API design rules |
| Team conventions | Framework-specific gotchas |
Tips
- Keep CLAUDE.md under 200 lines — use rules/ for overflow
- One rule per line is easier to maintain than paragraphs
- Include the concrete command, not just the concept
- Review promoted rules quarterly — remove what's no longer relevant
Related skills
- eas-app-stores – Curated upstream guidance for Eas App Stores; use when the workflow matches the user goal.
- eas-hosting – Curated upstream guidance for Eas Hosting; use when the workflow matches the user goal.
- eas-observe – Curated upstream guidance for Eas Observe; use when the workflow matches the user goal.
- eas-simulator – Curated upstream guidance for Eas Simulator; use when the workflow matches the user goal.
- prompt-governance – Use when managing prompts in production at scale: versioning prompts, running A/B tests…
- ecl-harness-engineer – Create or audit ECL Agent Harness infrastructure: AGENTS.md, change tracking, repository…
- effective-agent-skills – Author and review high-quality agent skills with triggers, progressive disclosure, and…
- subagent-driven-development – Execute plans via delegate_task subagents (2-stage review).