Skill: darwin-path-mode

Preconditions

• EVOLUTION_PATH.md exists at repo root (shipped with the template from v2.0).
• The OS has at least passed Rung 1 (bootstrap complete) — otherwise the only recommendation is "complete bootstrap first".
• The Darwin agent (.claude/agents/darwin.md) is available.

When to use

• The operator wants to see where they are on the evolution ladder and what's next.
• The operator hit a milestone (finished a domain, ingested a canon, ran first weekly ritual) and wants to know what compounds next.
• The operator has been at the same rung for a while and wants a nudge.

NOT for:

• Routine drift / housekeeping → use darwin-housekeeping.
• Deep governance with OS Health Report → use Darwin deep mode directly.
• Decision logging → use decision-log-entry.

Sequence

Step 1 — Invoke Darwin in path mode

Agent(darwin) with the following framing:
"Path mode — read EVOLUTION_PATH.md and evaluate where the operator currently sits.
Produce the canonical path-mode report (current settled rung + Done + Next 2-3 + Visible but distant + Not applicable yet).
End with the never-push line. Do not propose any rung the operator hasn't passed prerequisites for."

Step 2 — Surface the report

Pass Darwin's output through to the operator unmodified except for any sensitive references Walter would flag.

Step 3 — If the operator picks a rung

If the operator responds with "let's do rung N" or equivalent:

• Invoke the relevant downstream skill / agent for that specific rung. Examples:
  • Rung 1 → os-bootstrap
  • Rung 3 → decision-log-entry
  • Rung 4 → os-bootstrap-extend (domain micro-loop)
  • Rung 7 → os-bootstrap-extend (entity guardian micro-loop)
  • Rung 8 → ingest-content then create canon+self-audit pair manually
  • Rung 10 → memory-consolidate
  • Rung 12 → os-bootstrap-extend (orchestrator micro-loop)
• Update the relevant trackers when the rung completes (decision-log entry + agent registry + memory updates).

Step 4 — If the operator says "I'm settled at rung N"

• Write a decision-log-entry with type: meta recording the settled choice.
• Confirm Darwin will stop surfacing next-rung proposals until (a) explicit re-invocation, or (b) ≥ 30 days + new operating signal.

Anti-patterns

• Pushing. If the report becomes a sales pitch, the skill fails. The operator's freedom to stop is the design.
• Surfacing rungs whose prerequisites aren't met. Darwin must filter, not just enumerate.
• Ignoring the "Scardini's practice" mirror. The reference exists because pattern-only descriptions feel abstract; an embodied example calibrates expectations.
• Treating this as deep mode. Deep mode is drift detection. Path mode is direction proposal. Mixing them produces either drift-blind growth or growth-blind drift.

Example output

## 🪜 Where you are on the Evolution Path

**Current settled rung:** 7 · First entity guardian instantiated
**Phase:** Compounding

### ✅ Done
- Rung 1-5 · Foundations complete
- Rung 6 · Second domain (`finance/`) activated 2026-06-10
- Rung 7 · `family-guardian` instantiated 2026-06-15

### 🎯 Next 2-3 — ready + recommended
1. **Rung 8 · First canon ingested + self-audit paired**
   - Importance: high
   - Effort: M (~2 hours per canon, then re-check on the prescribed date)
   - Benefit vs current state: durable absorption of external knowledge — not just reading it once
   - Scardini's practice: first canon was Pyramid Principle (Minto); current shelf includes Dalio, Pfeffer, Kahneman, Taleb tetralogy, Ariely, Covey
2. **Rung 9 · Weekly governance ritual**
   - Importance: high
   - Effort: S to set up (~30 min), L to maintain (weekly forever)
   - Benefit vs current state: drift detection becomes automatic
   - Scardini's practice: Friday 17:00, Walter pressure-tests every proposal

### 🔮 Visible but distant
- Rung 11 · Concept cards — needs at least 2 ingested canons first

### 🚫 Not applicable yet
- Rung 12 · Orchestrator — requires 3+ named specialists in one domain

---

*You can settle anywhere on this ladder. Tell me the rung you want to stop at and I'll stop surfacing the next ones.*

Coherence with other skills

This skill is complementary to:

• os-bootstrap — bootstraps Rung 1.
• os-bootstrap-extend — operates Rungs 4, 6, 7, 12, 19, 20 (any new agent/domain/guardian/orchestrator/fallback).
• darwin-housekeeping — watchdog rhythm; runs daily independent of path mode.
• memory-consolidate — operates Rung 10 (auto-memory curation).
• ingest-content + manual canon authoring → Rungs 8, 13, 18.
• decision-log-entry → Rung 3 + meta entries for rung-settlement choices.

The skill is mutually exclusive with Darwin deep mode in the same turn — they answer different questions and merging them would dilute both.