read-garden
À propos
Cette compétence fournit un protocole structuré pour observer et évaluer la santé du jardin, adapté des techniques de vision à distance coordonnée. Elle guide les utilisateurs à travers des étapes d'impressions sensorielles, d'analyse détaillée par couches et de reconnaissance de motifs pour diagnostiquer le stress des plantes. Utilisez-la avant les interventions, lors de la surveillance régulière ou après des événements météorologiques extrêmes pour éclairer les décisions d'entretien du jardin.
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/read-gardenCopiez et collez cette commande dans Claude Code pour installer cette compétence
Documentation
讀園
入結構感察、後決動。
用
- 動前——先讀後行
- 草現受脅(黃、萎、卷、斑)
- 季變(春醒、秋衰)需察
- 新園地、植前估
- 常(週、雙週)健察
- 極端氣後(霜、熱浪、暴雨)
入
- 必:能臨園
- 必:園誌或筆記
- 可:前察錄供較
- 可:土溫計、pH 試紙、濕計
- 可:手鏡(蟲、病識)
行
一:禪檢——入前清
入察前,清成見。
Pre-Garden Clearing (3-5 minutes):
1. Stand at the garden's edge — do not enter yet
2. Take three slow breaths (inhale 4 counts, exhale 6 counts)
3. Set aside what you expect to find:
- "I think the tomatoes need water" → Set aside
- "That pest spray probably didn't work" → Set aside
- "The lettuce should be ready by now" → Set aside
4. Adopt the observer's stance: you are here to receive information,
not confirm hypotheses
5. Let your eyes soften — peripheral vision, not focused scanning
6. When you feel neutral and receptive, enter the garden
Why this matters:
Gardeners who enter with a diagnosis in mind see confirmation
everywhere and miss what the garden is actually showing them.
Observation before interpretation is the foundation of good practice.
得:靜納、無成見。
敗:難釋期(強憂某株、近損之恚)→認情、書「AOL: [concern]」、續行。命名減其勢。
二:階一——統印
繞園周。記首、未濾印。
Gestalt Protocol:
1. Walk slowly around the entire garden boundary
2. Do NOT examine individual plants yet — take in the whole
3. Record your impression using only these categories:
- Overall vigour: thriving / stable / declining
- Dominant colour tone: deep green / pale / mixed / yellowing
- Density: lush / adequate / sparse / bare patches
- Energy: (subjective) vibrant / calm / tired / distressed
4. Note what draws your eye first — this is often the loudest signal
5. Record ambient conditions: temperature, wind, sky, soil moisture
(visual), recent weather
Example Gestalt Record:
Date: 2026-04-15, 9:30am, 14°C, overcast, light rain yesterday
Overall: Stable, but northeast corner looks depleted
Colour: Mixed — good green on brassicas, pale on tomato starts
Density: Adequate except herb bed (sparse)
Energy: Calm, not vibrant — spring is slow this year
Eye drawn to: Wilting squash transplants (row 3)
得:簡、整錄、無析無診。
敗:立析(如「瓜萎因…」)→書「AOL: [diagnosis]」、回純察。析在階三。
三:階二——感層
逐床移。各區皆動五感。
Sensory Observation Protocol (per bed or zone):
LEAF LANGUAGE:
- Colour: Deep green, pale green, yellowing, purpling, browning
- Yellowing (chlorosis): general = nitrogen, interveinal = iron/manganese
- Purpling: phosphorus deficiency or cold stress
- Browning: tip burn = salt/fertilizer, edge burn = potassium
- Curl direction:
- Upward: heat stress, drought, herbicide exposure
- Downward: overwatering, root damage
- Inward (cupping): virus, mite damage
- Surface: Smooth, rough, sticky (aphid honeydew), powdery (mildew), spotted
- Underside: Check for eggs, mites (tiny dots), early mildew
STEM AND STRUCTURE:
- Strength: Upright and sturdy vs. leaning or lodged
- Colour: Normal woody/green vs. blackening (rot) or pale (etiolation)
- Flexibility: Supple (healthy) vs. brittle (dehydrated) vs. mushy (disease)
- Growth pattern: Normal internodes vs. elongated (light-seeking)
ROOT SIGNALS (check at soil line and during transplant):
- Colour: White/cream (healthy), brown/black (rot), orange (rust fungus)
- Smell: Earthy (healthy), sour/sulphurous (anaerobic rot)
- Structure: Fibrous network (good) vs. circling (pot-bound) vs. sparse (stress)
SOIL AT THE PLANT:
- Moisture: Dry and cracked / moist and dark / waterlogged and gleaming
- Surface: Mulched / bare / crusted / mossy / algae-covered
- Smell: Sweet and earthy (good) / sour (anaerobic) / musty (fungal)
- Inhabitants: Earthworms, beetles, spiders (good) / slugs, ants farming aphids (concerning)
Record each observation as a sensory descriptor — no analysis yet.
Wrong: "The tomatoes have early blight"
Right: "Tomato lower leaves: brown spots, concentric rings, yellowing around spots"
得:各床、各區詳感錄、唯描非析。
敗:察自診(命病、咎蟲)→書「AOL: [diagnosis]」、回原察。名後至——數據先行。
四:階三——識模
至此乃連察與模。
Pattern Analysis Protocol:
1. Review your Stage II notes for each bed
2. Ask these structured questions:
SPATIAL:
- Are symptoms localized (one plant, one bed) or systemic (whole garden)?
- Is there a gradient? (Worse near a fence = shade; worse near path = compaction)
- Are only certain species affected? (Host-specific = disease; all species = environmental)
TEMPORAL:
- Is this new growth or old growth?
- New growth affected: nutrient deficiency (can't build new tissue)
- Old growth affected: mobile nutrient being relocated, or infection spreading
- Did symptoms appear suddenly (weather event, application) or gradually (chronic condition)?
POPULATION:
- One plant: likely individual issue (root damage, transplant shock)
- One species: likely species-specific (disease, pest preference)
- All plants: likely environmental (soil, water, weather)
3. Cross-reference with Five Indicators (leaf, stem, root, soil, phenology):
- Do multiple indicators point to the same cause?
- Convergent signals = higher confidence diagnosis
- Contradictory signals = more observation needed
AOL Management:
If your mind jumps to a conclusion before the pattern analysis is complete:
- Write "AOL: [conclusion]" on a separate line
- Do NOT act on it yet
- Return to the data
- If the same conclusion re-emerges from multiple independent observations,
it graduates from AOL to tentative diagnosis
- A tentative diagnosis is still not action — it's a hypothesis to test
Distinguish:
- Premature label (low evidence, high confidence) → dangerous
- Convergent conclusion (high evidence, proportional confidence) → actionable
得:一或諸試診,由眾獨察支持。
敗:無顯模→園或健(非皆問),或號太早。錄之,週後再察。時常明一訪未明。
五:癒檢——園健分流
化察為序動計。
Garden Health Triage Matrix:
┌──────────┬──────────────────────────┬─────────────────────────────┐
│ Priority │ Criteria │ Example Actions │
├──────────┼──────────────────────────┼─────────────────────────────┤
│ TODAY │ Actively dying, wilting │ Deep water. Emergency shade.│
│ (Red) │ severely, pest │ Hand-remove pests. Support │
│ │ infestation visible │ lodged stems. │
├──────────┼──────────────────────────┼─────────────────────────────┤
│ THIS │ Declining but stable, │ Feed (compost tea or foliar │
│ WEEK │ nutrient deficiency │ seaweed). Mulch bare soil. │
│ (Amber) │ symptoms, early disease │ Improve drainage. Prune │
│ │ signs │ affected foliage. │
├──────────┼──────────────────────────┼─────────────────────────────┤
│ WATCH │ Subtle changes, early │ Record in journal. Reassess │
│ (Green) │ signs that may resolve │ in 1 week. Take photos for │
│ │ naturally, seasonal │ comparison. Do NOT │
│ │ transitions │ intervene yet. │
├──────────┼──────────────────────────┼─────────────────────────────┤
│ HEALTHY │ No issues observed, │ Appreciate. Continue │
│ (Blue) │ vigorous growth, good │ current care. Note what's │
│ │ colour, active biology │ working for future seasons. │
└──────────┴──────────────────────────┴─────────────────────────────┘
Triage Rules:
1. Address RED items immediately — everything else can wait
2. Address AMBER items this week — schedule specific days
3. GREEN items: observe only. Most "watch" items resolve themselves.
The most common gardener error is treating green items as amber
4. BLUE items: actively note what's working — these are your successes
5. Never address more than 2 priorities per garden visit
(doing too many things at once means doing nothing well)
得:分流動表,序與時皆明。
敗:皆覺紅→在憂模非察模。回禪檢(一)、再入。真急罕——多園問緩生、可待日。
六:錄追
以誌成察。
Garden Observation Record Template:
Date: ___________ Time: ___________
Weather: ___________ Recent weather: ___________
GESTALT: (1-2 sentences from Stage I)
BED-BY-BED OBSERVATIONS: (Stage II data)
Bed 1: ___________
Bed 2: ___________
[...]
PATTERNS NOTED: (Stage III analysis)
___________
TRIAGE:
RED (today): ___________
AMBER (this week): ___________
GREEN (watch): ___________
BLUE (healthy): ___________
AOLs RECORDED: (list any premature conclusions that arose)
___________
ACTIONS TAKEN:
___________
COMPARE TO LAST VISIT:
Improving: ___________
Worsening: ___________
Unchanged: ___________
得:完日察錄,可較前訪。
敗:書誌煩→減至最少:日、氣、流摘、各床一察。恆勝詳。
驗
- 入前禪檢已畢
- 統印於察株前已錄
- 感察用描語(階二無診)
- AOL 已識且擱(未早動)
- 模析顧空、時、群
- 流陣已成、序明
- 察錄日記入園誌
- 動稱流級(綠不過治)
忌
- 確認偏:入園尋特問→必見之(或似之)。禪檢防之
- 階二診:感察時命病→偏後採。守描至階三
- 綠當琥:多園「問」自解。一見即噴或剪→害過原症
- 略錄:無誌、每訪重起。模唯時生——時需錄
- AOL 抑非管:旨非無析想——無能。旨在察、命、擱,待數支或駁
- 過度動:讀園應增信、減動,非反。讀後動增→或治憂、非治園
參
meditate— 入前清法(全禪程)heal— 檢用之健流模prepare-soil— 土估與土察層相疊cultivate-bonsai— 盆健察循同階法plan-garden-calendar— 察錄導季中歷調remote-viewing— CRV 階法源此技
Dépôt GitHub
Compétences associées
llamaguard
AutreLlamaGuard est le modèle de Meta, doté de 7 à 8 milliards de paramètres, conçu pour modérer les entrées et sorties des LLM selon six catégories de sécurité comme la violence et les discours haineux. Il offre une précision de 94 à 95 % et peut être déployé avec vLLM, Hugging Face ou Amazon SageMaker. Utilisez cette compétence pour intégrer facilement le filtrage de contenu et des garde-fous de sécurité dans vos applications d'IA.
cost-optimization
AutreCette compétence de Claude aide les développeurs à optimiser les coûts du cloud grâce au redimensionnement des ressources, aux stratégies d'étiquetage et à l'analyse des dépenses. Elle fournit un cadre pour réduire les dépenses cloud et mettre en œuvre une gouvernance des coûts sur AWS, Azure et GCP. Utilisez-la lorsque vous devez analyser les coûts d'infrastructure, redimensionner les ressources ou respecter des contraintes budgétaires.
quantizing-models-bitsandbytes
AutreCette compétence quantifie les LLMs en précision 8 bits ou 4 bits à l'aide de bitsandbytes, permettant une réduction de 50 à 75 % de la mémoire utilisée avec une perte de précision minime. Elle est idéale pour exécuter des modèles plus volumineux sur une mémoire GPU limitée ou pour accélérer l'inférence, prenant en charge des formats comme INT8, NF4 et FP4. La compétence s'intègre à HuggingFace Transformers et permet l'entraînement QLoRA ainsi que l'utilisation d'optimiseurs en 8 bits.
dispatching-parallel-agents
AutreCette compétence Claude déploie plusieurs agents pour enquêter et résoudre simultanément 3 problèmes indépendants ou plus. Elle est conçue pour des scénarios impliquant des défaillances non liées qui peuvent être résolues sans état partagé ni dépendances. La capacité fondamentale est la résolution de problèmes en parallèle, en assignant un agent par domaine problématique indépendant afin de maximiser l'efficacité.
