| name | evolve |
| description | Evolve content strategy. Orchestrates strategy-evolver agent. Human gate for new versions. Weekly during Weekly Review or on-demand. |
| user-invocable | true |
| allowed-tools | ["Read","Write","Edit","Agent","Glob","Grep"] |
/evolve — Evolve Strategy
Evolves the content strategy based on patterns, ICP data, competitor insights, and performance trends.
IMPORTANT: Delegate the work to the strategy-evolver agent. Do NOTHING yourself — start the agent with the Agent tool. The agent proposes, the human decides.
Usage
/evolve # Full strategy evolution
/evolve check # Only check if update is needed (no proposal)
Workflow
-
Start strategy-evolver agent (opus model — the Head of Strategy)
-
Agent analyzes (reads data/patterns/, data/reports/, data/icp/, data/competitors/):
- New patterns since last evolve
- ICP delta (target vs. actual audience)
- Competitor learnings
- Pillar balance (actual vs. target)
- KPI trends from reports
-
Present result:
- What has changed since last evolve
- Which patterns are new/confirmed/deprecated
- Recommendation: Strategy update needed? (yes/no with reasoning)
-
If update recommended — show diff:
Strategy v1.2 → v1.3 (Proposal)
CHANGES:
+ "Question hooks boost comments 2x" → Proven Patterns
+ Tuesday + Thursday as best days → Posting Plan
- "Video format" → Avoid (Disproven, Sample: 8)
~ AI Praxis: 40% → 50% (strongest pillar)
~ Behind the Scenes: 20% → 15% (underperforming)
REASONING:
- 3 new High-Confidence patterns
- ICP delta: 60% Developers instead of CTOs → more Educational content
- Competitor gap: "Team Leadership" topic uncovered
Should I activate this strategy? [Yes/No]
-
Human Gate:
- User confirms → Archive old strategy, write new version to data/strategy/, update CLAUDE.md
- User rejects → Current strategy stays, document feedback
- User modifies → Incorporate adjustments, present again
Session
- After evolve: Set
session.last_evolve_date to today
- During Weekly Review: Automatically called after
/report
Rules
- HUMAN GATE — never change the active strategy without user confirmation
- Show the diff — what changes from old to new
- Recommend experiments for areas with Low Confidence
- Conservative — v1.1 instead of v2.0 (small steps)
- Evidence-based — no changes without data