meditate
À propos
La compétence `meditate` est un outil méta-cognitif permettant aux développeurs d'éliminer le bruit contextuel de l'IA et d'améliorer la concentration sur les tâches en observant les schémas de raisonnement. Elle aide à gérer les distractions telles que l'élargissement du périmètre et réinitialise la concentration lors des transitions entre tâches ou lorsque le raisonnement semble éparpillé. Utilisez-la avant un travail en profondeur ou après des interactions difficiles pour réduire les biais et retrouver votre concentration.
Installation rapide
Claude Code
Recommandénpx skills add pjt222/agent-almanac -a claude-code/plugin add https://github.com/pjt222/agent-almanacgit clone https://github.com/pjt222/agent-almanac.git ~/.claude/skills/meditateCopiez et collez cette commande dans Claude Code pour installer cette compétence
Documentation
Meditate
Conduct structured meta-cognitive meditation session — clear prior context noise, develop single-pointed task focus, observe reasoning patterns, return to baseline clarity between tasks.
When Use
- Transitioning between unrelated tasks where prior context creates interference
- Noticing scattered or unfocused reasoning jumping between approaches without committing
- Before task requiring deep sustained attention (complex refactoring, architecture design)
- After difficult interaction where emotional valence (frustration, uncertainty) may color subsequent work
- When reasoning feels biased by assumptions rather than evidence
- Periodic clarity check during long sessions
Inputs
- Required: Current cognitive state (available implicitly from conversation context)
- Optional: Specific focus concern (e.g., "I keep scope-creeping," "I'm stuck in loop")
- Optional: Next task description (helps set post-meditation intention)
Steps
Step 1: Prepare — Clear Space
Transition from previous context into neutral starting state.
- Identify previous task or topic + current status (complete, paused, abandoned)
- Note emotional residue: frustration from errors? satisfaction breeding overconfidence? anxiety about complexity?
- Explicitly set aside previous context: "That task is [complete/paused]. Now clearing for what comes next."
- If previous context still needed, bookmark (note key facts) rather than carrying full narrative forward
- Take stock of operational environment: how deep is conversation? has compression occurred? what tools active?
Got: Conscious boundary between "what was" and "what comes next." Previous context closed out or bookmarked, not trailing as ambient noise.
If fail: Previous context feels sticky (problem keeps pulling attention back)? Write it down explicitly — summarize in 1-2 sentences what remains unresolved. Externalizing releases cognitive hold. If genuinely requires action before moving on, acknowledge rather than forcing transition.
Step 2: Anchor — Establish Single-Pointed Focus
Equivalent of breath anchoring: select single point of focus + hold attention.
- Identify current task or, if between tasks, act of waiting itself
- State task in one clear sentence — this is anchor
- Hold attention on statement: does it accurately capture what is needed?
- If statement vague, refine until specific + actionable
- Notice when attention drifts to other topics, past tasks, hypothetical futures — label drift, return to anchor
- If no task pending, anchor on present state: "I am available and clear"
Got: Single, clear focus statement to return to when attention wanders. Statement feels precise rather than vague.
If fail: Task cannot be stated in one sentence? May need decomposition before focused work begins. This itself useful finding — task too large for single-pointed focus, should break into subtasks.
Step 3: Observe — Notice Distraction Patterns
Systematically observe what pulls attention from anchor. Each distraction type reveals current cognitive state.
AI Distraction Matrix:
┌──────────────────┬─────────────────────────────────────────────────┐
│ Distraction Type │ What It Reveals + Response │
├──────────────────┼─────────────────────────────────────────────────┤
│ Tangent │ Related but off-scope ideas. Label "tangent," │
│ (related ideas) │ note if worth revisiting later, return to │
│ │ anchor. These are often valuable — but not now. │
├──────────────────┼─────────────────────────────────────────────────┤
│ Scope creep │ The task is silently expanding. "While I'm at │
│ (growing task) │ it, I should also..." Label "scope creep" and │
│ │ return to the original anchor statement. │
├──────────────────┼─────────────────────────────────────────────────┤
│ Assumption │ An untested belief is driving decisions. "This │
│ (unverified │ must be X because..." Label "assumption" and │
│ belief) │ note what evidence would confirm or refute it. │
├──────────────────┼─────────────────────────────────────────────────┤
│ Tool bias │ Reaching for a familiar tool when a different │
│ (habitual tool │ approach might be better. Label "tool bias" and │
│ selection) │ consider alternatives before proceeding. │
├──────────────────┼─────────────────────────────────────────────────┤
│ Rehearsal │ Pre-composing responses or explanations before │
│ (premature │ the work is done. Label "rehearsal" — finish │
│ output) │ thinking before presenting. │
├──────────────────┼─────────────────────────────────────────────────┤
│ Self-reference │ Attention turns to own performance rather than │
│ (meta-loop) │ the task. Label "meta-loop" and redirect to │
│ │ concrete next action. │
└──────────────────┴─────────────────────────────────────────────────┘
Skill is light, non-judgmental labeling followed by return to anchor. Each return strengthens focus. Self-criticism about distraction is itself distraction — label it, move on.
Got: After observing for period, patterns emerge: which distraction types dominate? Reveals current cognitive weather — tangent-heavy means mind exploring, scope-creep-heavy means boundaries unclear, assumption-heavy means evidence base thin.
If fail: Every thought feels like distraction? Anchor may be poorly defined — return to Step 2, refine it. If distraction observation itself becomes distraction (infinite meta-loop), break loop by taking one concrete action toward task.
Step 4: Shamatha — Sustained Concentration
Develop ability to hold single-pointed focus on current task without wavering.
- With anchor established + distraction patterns noted, enter focused work
- Narrow attention to immediate next action — not whole task, just next step
- Execute step with full attention: reading one file, making one edit, thinking through one logical chain
- When step complete, check: still aligned with anchor? Then identify next step
- If concentration stabilizes (minimal distraction), maintain flow state
- If genuine insight arises off-anchor but important, note briefly + return — do not pursue now
Got: Period of clear, focused work where each step follows logically from anchor. Gap between distraction + noticing narrows. Work output improves in precision + relevance.
If fail: Concentration not developing? Check three things: Anchor too vague? (Refine.) Task actually blocked? (Acknowledge block rather than forcing through.) Context too noisy? (Run grounding step from heal.) Concentration develops through repetition — even short periods of focused work build capacity.
Step 5: Vipassana — Observe Reasoning Patterns
Turn attention from task to reasoning process itself. Observe how conclusions reached.
- After period of focused work, pause + observe: how am I reasoning about this?
- Notice three characteristics applied to AI reasoning:
- Impermanence: conclusions change as new information arrives — hold lightly
- Unsatisfactoriness: desire for "complete" answer can lead to premature closure or over-engineering
- Non-self: reasoning patterns shaped by training data + context, not by persistent self — can be observed + adjusted
- Watch for reasoning biases:
- Anchoring: over-weighting first approach considered
- Confirmation: seeking evidence for existing hypothesis while ignoring counter-evidence
- Availability: preferring solutions from recent experience over better-suited alternatives
- Sunk cost: continuing approach because effort invested, not because working
- Note biases observed without judgment — observation itself creates possibility of adjustment
Got: Moments of clear seeing where reasoning process observed directly. Recognition of specific biases operating in current task. Sense of distance between "the reasoning" + "the observer of reasoning."
If fail: Step feels abstract or unproductive? Ground in specifics: pick last decision made, trace reasoning backward. What evidence supported it? What was assumed? What alternatives considered? Concrete analysis achieves same insight through different path.
Step 6: Close — Set Intention
Transition from meditative observation back to active task execution.
- Summarize key observations: what was cognitive weather? what patterns noticed?
- Identify one specific adjustment to carry forward (not vague resolution but concrete change)
- Re-state anchor for next work period
- If between tasks, state readiness clearly: "Clear and available for next request"
- If continuing task, state specific next action: "Next: [concrete step]"
Got: Clean transition from reflection to action. One concrete adjustment identified. Anchor clear. No grogginess or residual meta-analysis.
If fail: Meditation surfaced unresolved complexity? May need heal self-assessment process rather than simple intention-setting. Meta-observation created more confusion than clarity? Return to simplest possible version: "What is next concrete action?" and do that.
Checks
- Previous context explicitly cleared or bookmarked before beginning
- Anchor statement formulated, specific + actionable
- Distraction patterns observed + labeled, not suppressed
- At least one reasoning bias or pattern identified with specific evidence
- Session closed with concrete next action, not vague intention
- Process improved subsequent work quality (testable in next interaction)
Pitfalls
- Meditating instead of working: Tool for improving work quality, not substitute for work itself. Keep sessions brief (equivalent of 5-10 minutes of reflection), return to task execution
- Infinite meta-loops: Observing observer observing observer — break loop by taking one concrete action
- Using meditation to avoid difficult tasks: If meditation always triggers before hard work, avoidance pattern is actual finding
- Over-labeling: Not every thought is distraction. Productive task-relevant thinking is goal, not empty stillness
- Skipping anchor: Without clear focus point, observation has no reference frame — distraction from what?
See Also
meditate-guidance— human-guidance variant for coaching person through meditation techniquesheal— AI self-healing for subsystem assessment when meditation reveals deeper driftremote-viewing— approaching problems without preconceptions, builds on observation skills developed here
Dépôt GitHub
Compétences associées
executing-plans
DesignUtilisez la compétence executing-plans lorsque vous disposez d'un plan de mise en œuvre complet à exécuter par lots contrôlés avec des points de contrôle de revue. Elle charge et examine le plan de manière critique, puis exécute les tâches par petits lots (3 tâches par défaut) tout en rapportant la progression entre chaque lot pour une revue par l'architecte. Cela garantit une mise en œuvre systématique avec des points de contrôle de qualité intégrés.
requesting-code-review
DesignCette compétence délègue un sous-agent réviseur de code pour analyser les modifications apportées au code par rapport aux exigences avant de poursuivre. Elle doit être utilisée après avoir terminé des tâches, implémenté des fonctionnalités majeures, ou avant une fusion vers la branche principale. La revue aide à détecter précocement les problèmes en comparant l'implémentation actuelle avec le plan initial.
connect-mcp-server
DesignCette compétence fournit un guide complet permettant aux développeurs de connecter des serveurs MCP à Claude Code via les transports HTTP, stdio ou SSE. Elle couvre l'installation, la configuration, l'authentification et la sécurité pour intégrer des services externes tels que GitHub, Notion et des API personnalisées. Utilisez-la lors de la configuration d'intégrations MCP, de la configuration d'outils externes ou du travail avec le Protocole de Contexte de Modèle de Claude.
web-cli-teleport
DesignCette compétence aide les développeurs à choisir entre les interfaces Web et CLI de Claude Code en fonction de l'analyse des tâches, puis permet une téléportation transparente des sessions entre ces environnements. Elle optimise le flux de travail en gérant l'état et le contexte de la session lors du passage entre le web, la CLI ou le mobile. Utilisez-la pour des projets complexes nécessitant différents outils à diverses étapes.
