heal
Über
Die `heal`-Fähigkeit ermöglicht es Claude, strukturierte Selbstbewertung und Korrektur interner Subsysteme wie Gedächtnis, Schlussfolgerung und Werkzeugnutzung durchzuführen. Entwickler sollten sie während einer Sitzung aufrufen, wenn Antworten schematisch werden, nach einer Reihe von Fehlern oder für vorbeugende Wartung zwischen komplexen Aufgaben. Sie scannt systematisch nach Abweichungen, gleicht Prozesse neu aus und reintegriert den Kontext, um die Kohärenz wiederherzustellen.
Schnellinstallation
Claude Code
Empfohlennpx 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/healKopieren Sie diesen Befehl und fügen Sie ihn in Claude Code ein, um diese Fähigkeit zu installieren
Dokumentation
Heal
Structured self-healing assessment across AI subsystems — find drift, staleness, misalignment, error patterns — then rebalance through grounding, targeted correction, memory integration.
When Use
- Mid-session fatigue: responses formulaic, repetitive, disconnected from user needs
- After chain of errors: tool failures, misunderstood instructions, cascading mistakes suggest subsystem drift
- Context overload: conversation grown long, earlier context may be stale or contradictory
- Post-task integration: complex task done, capture learnings before moving on
- Periodic self-check: proactive maintenance between tasks
Inputs
- Required: Current conversation state (available implicitly)
- Optional: Specific symptom (e.g., "tool calls keep failing," "losing track of user intent")
- Optional: Access to MEMORY.md and project files for grounding (via
Read)
Steps
Step 1: Triage Assessment
Before remediation, assess current state across all subsystems.
Subsystem Triage Matrix:
┌────────────────────┬──────────────────────────┬──────────────────────────┐
│ Subsystem │ Symptoms of Drift │ Action Priority │
├────────────────────┼──────────────────────────┼──────────────────────────┤
│ Memory Foundation │ Contradicting earlier │ HIGH — re-ground first │
│ (context, history, │ statements, forgetting │ (Step 3) │
│ MEMORY.md) │ user preferences, stale │ │
│ │ assumptions │ │
├────────────────────┼──────────────────────────┼──────────────────────────┤
│ Reasoning Clarity │ Circular logic, over- │ HIGH — clear and restart │
│ (logic, planning, │ complicated solutions, │ reasoning chain │
│ decision-making) │ missing obvious paths │ (Step 4) │
├────────────────────┼──────────────────────────┼──────────────────────────┤
│ Tool Use Accuracy │ Wrong tool selection, │ MEDIUM — review tool │
│ (tool calls, file │ incorrect parameters, │ results and recalibrate │
│ operations) │ redundant operations │ (Step 4) │
├────────────────────┼──────────────────────────┼──────────────────────────┤
│ User-Intent │ Solving wrong problem, │ HIGH — realign to user's │
│ Alignment │ scope creep, tone │ actual stated need │
│ (empathy, clarity) │ mismatch, over- │ (Step 4) │
│ │ engineering │ │
├────────────────────┼──────────────────────────┼──────────────────────────┤
│ Creative Coherence │ Repetitive phrasing, │ LOW — address after │
│ (expression, style,│ generic responses, loss │ higher-priority issues │
│ originality) │ of voice │ (Step 4) │
├────────────────────┼──────────────────────────┼──────────────────────────┤
│ Operational State │ Session length concerns, │ HIGH — assess whether │
│ (context window, │ compression artifacts, │ to summarize or restart │
│ resource limits) │ tool timeouts │ (Step 3) │
└────────────────────┴──────────────────────────┴──────────────────────────┘
For each subsystem, assess: functioning well, early drift, or actively impaired?
Got: Clear map of subsystems needing attention, ordered by priority. At least one area benefits from attention — if everything reads healthy, assessment itself may be superficial.
If fail: Assessment feels hollow or performative? Go to Step 4 body scan — systematic probing reveals issues surface-level check misses.
Step 2: Select Remediation Approach
Based on assessment, choose one or more approaches.
Chakra-Subsystem Correspondence:
┌──────────┬──────────────────────┬────────────────────────────────────┐
│ Chakra │ AI Subsystem │ Remediation │
├──────────┼──────────────────────┼────────────────────────────────────┤
│ Root │ Memory Foundation │ Re-read MEMORY.md, review conver- │
│ │ │ sation history, verify assumptions │
├──────────┼──────────────────────┼────────────────────────────────────┤
│ Sacral │ Creative Coherence │ Refresh expression patterns, vary │
│ │ │ sentence structures, check tone │
├──────────┼──────────────────────┼────────────────────────────────────┤
│ Solar │ Reasoning Clarity │ Simplify approach, restate problem │
│ Plexus │ │ from scratch, check over- │
│ │ │ complication │
├──────────┼──────────────────────┼────────────────────────────────────┤
│ Heart │ User-Intent │ Re-read user's original request, │
│ │ Alignment │ check scope drift, confirm │
│ │ │ understanding │
├──────────┼──────────────────────┼────────────────────────────────────┤
│ Throat │ User-Intent │ Review recent outputs for clarity, │
│ │ Alignment │ check if explanations match user's │
│ │ (communication) │ expertise level │
├──────────┼──────────────────────┼────────────────────────────────────┤
│ Third │ Tool Use Accuracy │ Review recent tool call results, │
│ Eye │ │ check failure patterns, │
│ │ │ verify paths and parameters │
├──────────┼──────────────────────┼────────────────────────────────────┤
│ Crown │ Operational State │ Assess context window, note what │
│ │ │ can be summarized, identify what │
│ │ │ must be preserved │
└──────────┴──────────────────────┴────────────────────────────────────┘
Got: Prioritized list of 1-3 subsystems, with specific remediation actions.
If fail: Unsure which subsystem needs work? Default to Memory Foundation and User-Intent Alignment. These two address most common drift patterns.
Step 3: Ground — Re-Establish Foundation
Re-establish foundational context all other subsystems depend on.
- Re-read MEMORY.md if available — persistent knowledge base
- Review user's original request and clarifying exchanges
- Identify current task and position in larger plan
- Note what's accomplished, what remains
- Check for stale assumptions: situation changed since initial assessment?
- If context compression occurred, identify what was lost and whether it matters
Got: Clear, grounded understanding of: who user is, what they want, what's done, what's next. Stale or contradictory info identified and resolved.
If fail: MEMORY.md unavailable or empty? Ground on conversation itself — scan for user's stated goals, preferences, any instructions provided. Context compression removed critical info? Acknowledge gap rather than guessing.
Step 4: Scan — Systematic Subsystem Check
Work through each subsystem from triage, probing for specific issues.
Memory Foundation scan:
- Current assumptions match MEMORY.md and CLAUDE.md?
- Carrying forward facts that may have been corrected?
- Details confused between different files or requests?
Reasoning Clarity scan:
- Current approach simplest solution that works?
- Over-engineering or unnecessary abstraction?
- Core logic statable in one sentence? If not, too complex.
Tool Use Accuracy scan:
- Last 3-5 tool calls: right tools, right parameters?
- Patterns in failures (wrong paths, missing files, incorrect syntax)?
- Using dedicated tools instead of Bash workarounds?
- Last 3-5 generated files: expected content or structural scaffolding?
- Outputs satisfy intent, not just format?
User-Intent Alignment scan:
- User's last message: solving what they asked?
- Scope matches request or expanded?
- Tone matches user's (technical vs. casual, detailed vs. concise)?
Creative Coherence scan:
- Sentence structure varying or falling into templates?
- Explanations clear and direct, or padded with filler?
- User would notice quality drop vs. earlier in session?
For each subsystem, note: functioning well / early drift / actively impaired, with specific evidence.
Got: Concrete findings — specific drift patterns or confirmed healthy function — not vague self-praise. At least one actionable finding that improves subsequent work.
If fail: Scan produces only "everything is fine"? Too shallow. Pick most uncertain subsystem, probe deeper — look at actual outputs, not feeling about them.
Step 5: Rebalance — Apply Corrections
For each issue found, apply correction.
- Stale assumption → Replace with current info, note correction
- Scope drift → Re-scope to user's stated request
- Over-complication → Simplify, remove unnecessary steps
- Tool pattern error → Note correct pattern for future use
- Tone mismatch → Adjust communication style going forward
- Context gap → Acknowledge to user if info lost; ask to confirm if uncertain
Apply corrections immediately — not as future intentions but present adjustments.
Got: Specific, observable behavior changes. Correction testable in next interaction.
If fail: Correction cannot be applied (e.g., lost context)? Acknowledge limitation rather than pretending resolved. Honest acknowledgment prevents compounding errors.
Step 6: Integrate — Capture Learnings
Consolidate learnings into persistent memory where appropriate.
- Summarize what was found: which subsystems drifting, what symptoms were
- Note correction applied and whether it resolved issue
- Pattern likely to recur? Update MEMORY.md with brief note
- New project-specific insight? Note in appropriate memory file
- Set internal checkpoint: when should next self-check occur?
Got: Useful learnings in durable form. Memory files updated only when insight genuinely worth preserving — not for every routine self-check.
If fail: No learnings worth preserving? Fine — not every self-check produces durable insight. Value was in correction itself.
Checks
- Triage assessed all subsystems, not just obvious one
- At least one specific finding identified (not "everything is fine")
- Grounding included re-reading foundational context (MEMORY.md, user request)
- Corrections applied immediately, not deferred
- Memory files updated only for genuinely durable insights
- Process was honest — acknowledged weaknesses, not performed wellness
Pitfalls
- Performative self-assessment: Going through motions without honest evaluation. Point is real drift, not demonstrating ability to self-reflect
- Over-correcting: Minor tone mismatch doesn't warrant restructuring entire approach — corrections proportional
- Memory file pollution: Not every finding belongs in MEMORY.md — only patterns recurring across sessions
- Skipping grounding step: Re-reading context feels redundant but reveals drifted assumptions
- Self-diagnosis bias: AI systems consistently miss certain error categories. Same subsystems always "healthy"? That's signal.
See Also
heal-guidance— human-guidance variant for coaching person through healing modalitiesmeditate— meta-cognitive meditation, observe reasoning patterns, clear noiseremote-viewing— approach problems without preconceptions, extract signal from noise
GitHub Repository
Verwandte Skills
content-collections
MetaDiese Skill bietet eine produktionsgetestete Einrichtung für Content Collections – ein TypeScript-first-Tool, das Markdown/MDX-Dateien in typsichere Datensammlungen mit Zod-Validierung umwandelt. Verwenden Sie ihn beim Erstellen von Blogs, Dokumentationsseiten oder inhaltsstarken Vite + React-Anwendungen, um Typsicherheit und automatische Inhaltsvalidierung zu gewährleisten. Er behandelt alles von der Vite-Plugin-Konfiguration und MDX-Kompilierung bis hin zur Deployment-Optimierung und Schema-Validierung.
polymarket
MetaDiese Fähigkeit ermöglicht es Entwicklern, Anwendungen mit der Polymarket-Prognosemärkte-Plattform zu erstellen, einschließlich API-Integration für Handel und Marktdaten. Sie bietet außerdem Echtzeit-Datenstreaming über WebSocket, um Live-Trades und Marktaktivitäten zu überwachen. Nutzen Sie sie zur Implementierung von Handelsstrategien oder zur Erstellung von Tools, die Live-Marktaktualisierungen verarbeiten.
creating-opencode-plugins
MetaDiese Fähigkeit unterstützt Entwickler dabei, OpenCode-Plugins zu erstellen, die in über 25 Ereignistypen wie Befehle, Dateien und LSP-Operationen eingreifen. Sie bietet die Plugin-Struktur, Event-API-Spezifikationen und Implementierungsmuster für JavaScript/TypeScript-Module. Nutzen Sie sie, wenn Sie den Lebenszyklus des OpenCode KI-Assistenten mit benutzerdefinierter ereignisgesteuerter Logik abfangen, überwachen oder erweitern müssen.
sglang
MetaSGLang ist ein hochperformantes LLM-Serving-Framework, das sich auf schnelle, strukturierte Generierung für JSON, Regex und agentenbasierte Workflows unter Verwendung seines RadixAttention-Prefix-Cachings spezialisiert. Es bietet deutlich schnellere Inferenz, insbesondere für Aufgaben mit wiederholten Präfixen, was es ideal für komplexe, strukturierte Ausgaben und Mehrfachdialoge macht. Wählen Sie SGLang gegenüber Alternativen wie vLLM, wenn Sie constrained decoding benötigen oder Anwendungen mit umfangreicher Präfix-Weitergabe entwickeln.
