beam-ai-team/learn-skills

[TARGET] Build-First Onboarding (CONTEXTUAL SUGGESTION)

This is a learning skill. With Build-First onboarding, suggest AFTER first task execution, not before.

Auto-Complete via Checkpoint

# In user-config.yaml
build_first.checkpoints.first_task_executed: true  → learn_skills auto-completed!

When user completes their first section in execute-project, learn_skills is automatically marked complete via auto_complete_map. They learned by DOING.

When to Proactively Suggest (Pattern Detection)

Check both:

• learning_tracker.completed.learn_skills - explicit completion
• build_first.checkpoints.first_task_executed - auto-completion

PROACTIVELY SUGGEST when user:

1. Pattern detected - does something similar to previous work ("I notice you've done this type of task before...")
2. Creates multiple similar builds (report-jan, report-feb anti-pattern)
3. Describes repeating work patterns ("every week I...", "I always have to...")
4. Explicitly asks to learn about skills/automation

Suggestion Pattern (After Pattern Detection):

💡 I notice you've done similar work before. Want to understand how to
turn repeating workflows into Skills?

'learn skills' covers:
- The 3-criteria skill-worthiness framework
- How skills are structured
- How AI triggers skills automatically

Say 'learn skills' for the deep-dive (10 min), or continue working.

Anti-Pattern Detection:

If user creates: report-jan, report-feb, report-mar...
→ "I notice you're creating similar items. This is a perfect use case for
   a SKILL instead of multiple projects. Want to 'learn skills' to understand
   how to capture this as a reusable workflow?"

DO NOT suggest if:

• User is working on their FIRST build/task (let them DO first)
• User explicitly says "skip" or dismisses
• User is mid-task and focused
• No repeating pattern detected

---

Learn Skills

Teach how to identify skill-worthy workflows and create effective skills.

Purpose

Help user understand what makes something skill-worthy, how skills are structured, and how skill triggering works. Includes hands-on practice identifying their own workflows.

Time Estimate: 10-12 minutes

---

Workflow

Step 1: Concrete Examples

[OK] SKILLS (repeating workflows):
- Weekly status report (same format weekly)
- Qualify sales lead (same questions each time)
- Process expense reports (same steps)

[ERROR] NOT SKILLS (one-time):
- Research competitor Acme (one-time)
- Build Q1 marketing plan (one-time)

Key question: Will I do this AGAIN?

---

Step 2: Skill-Worthiness Framework

Three questions:

1. FREQUENCY: 2+ times per month?
   YES → keep evaluating

2. REPEATABILITY: Steps mostly the same?
   YES → keep evaluating

3. VALUE: Saves >5 minutes per execution?
   YES → Create a skill!

ALL 3 YES = Skill-worthy
ANY NO = Just do it manually

---

Step 3: Skill Structure

[DIR] weekly-status-report/
├── SKILL.md       # Instructions + triggers
├── references/    # Documentation (optional)
├── scripts/       # Automation (optional)
└── assets/        # Templates (optional)

---

Step 4: How Triggering Works

AI checks your message against ALL skill descriptions.
Match found = skill loads.

Example description:
"Load when user says 'status report', 'weekly update',
 'progress summary'"

ANY of these triggers it:
• "Generate my status report"
• "Weekly update please"
• "Progress summary"

---

Step 5: Practice

Ask: "What did you do last week that you'll probably do again?"

For each: apply 3-criteria framework, brainstorm trigger phrases.

---

Step 6: How to Create

To create a skill, say:
• "create skill for [workflow]"
• "new skill: [name]"

YOUR skills go in 01-skills/ (prioritized!)
SYSTEM skills in 00-system/skills/ (in the public catalog these map under skills/… in beam-next-skills)

---

Step 7: Finalize

Actions (MUST complete all):

1. Mark skill complete in user-config.yaml:

   learning_tracker:
     completed:
       learn_skills: true  # ADD THIS LINE
   

2. Display completion:

   [OK] Learn Skills Complete!
   
   You now understand:
   • Skills = reusable workflows (do AGAIN → skill)
   • 3-criteria framework (Frequency + Repeatability + Value)
   • Skill structure (SKILL.md + optional references/scripts)
   • Trigger mechanism (keywords in description)
   
   Next steps:
   • 'create skill' - Create your first skill
   • 'learn projects' - Learn about temporal work
   • 'learn beam next' - System mastery
   

3. Session ending tip:

   💡 When you're done, open a NEW chat for your next topic.
   

---

Success Criteria

• [ ] User understands skill vs build distinction
• [ ] User can apply 3-criteria skill-worthiness framework
• [ ] User knows skill folder structure
• [ ] User understands trigger mechanism
• [ ] User identified at least one potential skill from their work
• [ ] learning_tracker.completed.learn_skills: true in user-config.yaml