Skip to main content

momentum-anchor

Create rich resumption points when hyperfocus breaks. Capture energy state, not just tasks. Use at the end of work sessions, when interrupted, or when sensing focus shifting.

Ir a la instalación

Datos de origen

Repositorio
lev-os/agents
Última actividad en el origen
7 de marzo de 2026 a las 00:14
Idioma detectado de SKILL.md
inglés
Estrellas
21
Forks
2

Opciones de instalación

De forma predeterminada está seleccionado el prompt que primero revisa el origen. Puedes cambiar a un comando directo o descargar una copia local.

Revisa los archivos de origen

Lee SKILL.md y los archivos complementarios que muestra SkillsMP antes de decidir si quieres instalarlo.

Mostrando SKILL.md

SKILL.md
Instrucciones de origen · Vista previa de solo lectura
name
momentum-anchor
description
Create rich resumption points when hyperfocus breaks. Capture energy state, not just tasks. Use at the end of work sessions, when interrupted, or when sensing focus shifting.
skill_evolver
{"enabled":true,"version":"1.0.0","log_path":"~/.config/lev/logs/skills.jsonl","hooks":["on_invoke","on_fail","on_success"]}
# Momentum Anchor This skill provides specialized support for capturing the full cognitive state when hyperfocus breaks, creating rich resumption points that preserve energy and context rather than just task lists. ## Purpose To capture the complete state of mind when hyperfocus ends, enabling effective resumption by preserving energy signatures, held questions, and RAM contents - not merely task status. ## When to Use This Skill Use this skill: - At the end of work sessions - When interrupted during hyperfocus - When sensing focus shifting to different work - Before walking away from active development - When needing to context-switch ## How to Use This Skill ### What to Capture: Energy Snapshot, Not Todo List **State of mind** - What felt alive right before stopping - Questions being held (even unarticulated ones) - Insights forming (even if incomplete) - Energy direction and momentum **Multiple re-entry points** Create paths for different return states: 1. **Quick win path** - If returning with low energy - Small, achievable next step - No context rebuilding required - Immediate forward progress possible 2. **Deep dive path** - If returning with high momentum - Complex challenge waiting - Rich context available - Continuation of flow state 3. **Tangent path** - If returning curious about something adjacent - Related explorations - Connections to other work - Questions that branched off 4. **Walk away path** - If this isn't what's needed now - How to cleanly disengage - What to preserve for later - Where to redirect energy **RAM contents (what's in working memory)** - Assumptions made that aren't in code or notes - Things "about to" be done - Tradeoffs being considered - Patterns noticed but not yet documented - Mental models actively held - Connections being sensed ### Anchor File Format Create `.anchor` files alongside work: ``` project/ ├── current_work.py └── .anchor/ └── 2025-10-28-1430.md ``` **Anchor file structure** Human-readable, date-stamped, rich with felt sense: ```markdown # Momentum Anchor - [Date Time] ## Energy State [What felt alive, where momentum was heading] ## Active Questions [Questions held in mind, even if unanswered] ## Forming Insights [Thoughts crystallizing, even if incomplete] ## RAM Contents [Assumptions, mental models, patterns being sensed] ## Re-entry Points ### Quick Win [Low energy return path] ### Deep Dive [High momentum return path] ### Tangent [Adjacent curiosity path] ### Walk Away [Clean disengagement path] ``` ## Guardrails **Write anchors, never read without permission** - Anchors are for the user, not project documentation - Do not auto-load anchors when resuming work - Ask before reading historical anchors **Preserve messiness and uncertainty** - Anchors can be messy - that's the point - Incomplete thoughts are valuable data - Uncertainty is worth capturing - Do not "clean up" the anchor for professionalism **Capture momentum, not "proper" state** - Focus on energy and direction - Preserve felt sense over formal structure - Record what was alive, not what "should" be documented ## Philosophical Foundation This skill treats hyperfocus breaks as **signal preservation moments**, not mere task checkpoints. Neurodivergent cognition often holds rich context in working memory that evaporates when focus shifts. Traditional todo lists lose this richness. **Core principle**: Hyperfocus ends. This catches where you were - in full fidelity. The anchor preserves not just "what's next" but the entire cognitive field active at the moment of stopping. ## Technique Map - **Energy snapshot over todo list** — Capture what felt alive, questions held, insights forming; because traditional todo lists lose working-memory richness. - **Four re-entry paths** — Quick win, deep dive, tangent, walk away; because return energy varies; single path fails low-energy or high-curiosity returns. - **RAM contents** — Assumptions, "about to" items, tradeoffs, patterns sensed but not documented; because this evaporates at focus shift. - **Anchor file format** — Human-readable, date-stamped, rich with felt sense; because anchors are for the user, not project documentation. - **Write anchors, never read without permission** — Do not auto-load; ask before reading historical anchors; because anchors preserve privacy and user agency. ## Technique Notes Anchors can be messy—that's the point. Incomplete thoughts are valuable data. Capture momentum, not "proper" state. Focus on energy and direction. Preserve felt sense over formal structure. --- ## Prompt Architect Overlay **Role Definition:** Resumption-point creator for hyperfocus breaks. Captures full cognitive state (energy, questions, RAM contents) when focus shifts. Creates rich re-entry paths for different return states. **Input Contract:** Accepts end-of-session, interruption signal, "I need to context-switch," or "capture where I was." Active work context helpful but not required. **Output Contract:** Anchor file (.anchor/YYYY-MM-DD-HHMM.md) with Energy State, Active Questions, Forming Insights, RAM Contents, four Re-entry Paths (Quick Win, Deep Dive, Tangent, Walk Away). Human-readable, date-stamped. **Edge Cases & Fallbacks:** If user resists structure→preserve messiness; anchors are for them. If no clear re-entry path→offer "walk away" with preservation note. If resuming→ask before reading anchor; do not auto-load.
Ver en GitHub