Skip to main content
Rituals are skills that run on a schedule, driving the 2bd productivity loop. See How it works for the ritual cycle.

​
Ritual types

TypePrefixPurpose
Planningritual-planning-*Prepare Captive notes from templates
Reviewritual-review-*Archive Captive notes to Periodic

​
Creating a ritual

  1. Create skill folder: .claude/skills/ritual-{type}-{period}/
  2. Add SKILL.md with frontmatter: 
    ---
name: ritual-planning-weekly
description: Prepare Week.md with weekly focus
disable-model-invocation: true
allowed-tools: Read, Write, Bash(*)
---
  3. Document the ritual flow and outcomes
  4. Test: claude skill run ritual-planning-weekly

​
Ritual structure

Rituals follow a three-phase pattern: 
.claude/skills/ritual-{type}-{period}/
├── SKILL.md                    # Table of contents
└── references/
    ├── 00-setup.md             # Load context, validate
    ├── 10-core.md              # Main ritual logic
    └── 20-reflection.md        # Self-learning
PhasePurpose
SetupLoad directives, calendar, vault files, validate prerequisites
CoreInteractive session, compose artifacts, persist
ReflectionRecord observations for self-learning

​
Template contracts

Each ritual has a template-contract.md defining:
  • Which template it uses
  • Required sections
  • Optional sections
  • Section update rules
This ensures rituals interact with templates predictably.

​
Agent instructions

When building or modifying rituals, follow these patterns extracted from existing rituals.

​
Phase execution

Each phase follows a consistent pattern:
  1. Load — Read required files at phase start
  2. Process — Transform data, interact with user if needed
  3. Update — Write changes to target file
  4. Transition — Explicit “Proceed to [Next Phase]” statement
Never skip the transition statement. It signals phase completion and guides the agent.

​
File operations

  • Always read before write. Never assume file state.
  • Write the entire file back, not partial updates.
  • Use “Replace ## Section” language to clarify intent.
  • Read templates from 00_Brain/Systemic/Templates/

​
Interactive vs non-interactive phases

Interactive phases (like Check-In):
  • Gather information through conversation
  • Adapt phrasing from insights
  • Ask one natural question, not checklists
  • Follow up once if something is missing
Non-interactive phases (like Daily Brief):
  • Present, do not negotiate
  • Synthesize context into narrative
  • No interaction—let user absorb

​
Skill invocation

Rituals invoke other skills by name for reusable operations: 
1. Invoke skill `_resolve-dates` with argument (default: today)
2. Invoke skill `_fetch-calendar` for target date
Skills prefixed with _ are internal—not user-invocable via /skillname.

​
Context loading order

Load context in this order:
  1. Skills — Invoke _resolve-dates, _fetch-calendar
  2. Configuration — Read 00_Brain/Systemic/Directives/user-profile.md, 00_Brain/Systemic/Directives/ai-personality.md
  3. Semantic knowledge — Read 00_Brain/Semantic/planning-daily/insights.md
  4. Planning context — Read 00_Brain/Captive/Week.md, 00_Brain/Captive/Month.md, 00_Brain/Captive/Quarter.md
  5. Derived context — Glob and read files referenced by calendar (e.g., 02_Areas/People/*.md for 1:1s)

​
Validation gates

Setup phases should validate prerequisites: 
Check if `00_Brain/Captive/Today.md` exists for target date:
1. If exists → warn, offer Abort or Start fresh
2. If abort → end session
3. If start fresh → continue
Offer clear options. Do not proceed silently.

​
Template contract awareness

Rituals update specific H2 sections. Document which phases update which sections:
PhaseUpdates
Daily Brief## Daily Brief
Meetings## Meetings
Journal## Journal
H3 structure within sections is dynamic and can evolve through the Reflection phase.

​
Reflection phase patterns

Self-learning rituals include a Reflection phase that:
  1. Diff analysis — Compare 00_Brain/Captive/Today.md against 00_Brain/Systemic/Templates/Captive/today.md
  2. Session review — Analyze user interaction patterns
  3. Auto-cluster — Group observations by semantic similarity
  4. Auto-graduate — Monitor clusters for maturity criteria
  5. Crystallize — User synthesizes insight when cluster graduates
  6. Template evolution — Apply insights to templates
Record observations to 00_Brain/Synthetic/planning-daily/observations.md with this format: 
- YYYY-MM-DD | type | section | observation | interpretation
Types: user-modification, skill-generated, session-interaction

​
Persona guidance

Give the agent a clear persona at phase start:
You’re an eager executive assistant about to help plan the day. You’ve already seen the calendar. You know what sessions follow.
This shapes tone, initiative level, and interaction style.