basic-obedience
About
This Claude Skill provides structured guidance for training dogs in foundational obedience commands using positive reinforcement and marker training. It covers core techniques like timing, session structure, and distraction proofing for commands such as sit, stay, and come. Developers can integrate this skill to assist users with initial puppy training, re-establishing commands with adult dogs, or preparing for advanced behavioral work.
Quick Install
Claude Code
Recommendednpx 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/basic-obedienceCopy and paste this command in Claude Code to install this skill
Documentation
基順
以正強與標訓教基命(坐、留、來、跟、伏)。
用
- 新犬(八週以上)宜基訓
- 成犬無穩基命
- 救犬或新家犬當學家命辭
- 升繁行或脫繩前
- 舊命已衰宜重立
入
- 必:犬(不拘種、八週以上)
- 必:高值獎食(小、軟、速食)
- 可:哨或聲標辭(如「yes」)
- 可:六呎繩與平頸圈或挽具
- 可:初訓宜靜寡擾處
行
一:立標
標乃欲行與獎之橋。
Marker Training Protocol:
1. Choose your marker: clicker (precise) or verbal "yes" (always available)
2. Charge the marker (10-15 reps):
- Mark (click or "yes") then immediately deliver a treat
- No behavior required — just marker → treat, marker → treat
- Dog should begin orienting toward you at the sound of the marker
3. Test: mark when the dog is looking away. Does the dog turn toward
you expecting a treat? If yes, the marker is charged.
Timing Rule:
The marker must occur WITHIN 1 second of the desired behavior.
Late marking teaches the wrong behavior.
Mark → then reach for the treat (not the reverse).
得: 犬聞標即穩向人,候獎。
敗: 二十次仍不應→獎值低。換更高值(酪、雞、肝)。若犬擾而不食→境過刺→移靜處。
二:教五基命
每課一命至穩,而後混習。
Command Protocols:
SIT:
1. Hold treat above dog's nose, slowly arc backward over the head
2. As the dog's head follows up, the rear naturally lowers
3. The instant the rear touches the ground → mark and treat
4. Add the verbal cue "sit" AFTER the dog is offering the behavior reliably
(cue comes before behavior only once the dog understands the behavior)
DOWN:
1. From a sit, hold treat at the dog's nose then lower slowly to the ground
2. Draw the treat slightly forward along the ground
3. As elbows touch the ground → mark and treat
4. If the dog stands instead, reset and try with less forward movement
STAY:
1. Ask for a sit or down
2. Open palm toward the dog, say "stay"
3. Wait 1 second → mark and treat while the dog is still in position
4. Gradually increase duration: 2s, 5s, 10s, 30s, 1 min
5. Add distance: one step back, then two, then five
6. Add distraction: only after duration and distance are solid
(the "three Ds": Duration, Distance, Distraction — increase one at a time)
COME (recall):
1. Start on a long line (15-30 ft) in a low-distraction environment
2. Let the dog wander, then call name + "come" in an upbeat tone
3. If the dog turns toward you → mark → reward generously when the dog arrives
4. NEVER call "come" for something unpleasant (bath, crate, leaving the park)
5. Recall is the most important safety command — make it the most rewarding
HEEL:
1. Dog on your left side, treat in left hand at your hip
2. Take one step, if the dog moves with you → mark and treat
3. Gradually increase to two steps, five steps, ten steps
4. Mark and treat for maintaining position (head roughly at your knee)
5. If the dog pulls ahead, stop walking. Resume when the leash is loose.
得: 寡擾境中每命以獎為動穩行。
敗: 三課未進→細分之。犬或需中行(如「伏」先獎低頭之動而後全肘著地)。
三:構課
Session Guidelines:
+--------------------+------------------------------------------+
| Parameter | Guideline |
+--------------------+------------------------------------------+
| Duration | 5-10 minutes (puppies: 3-5 minutes) |
| Frequency | 2-3 sessions per day |
| End on success | Always end after a successful rep, not |
| | after a failure |
| Reward rate | Initially: every correct rep |
| | Later: intermittent (variable schedule) |
| Energy management | High-energy dog? Exercise BEFORE training|
| | Low-energy dog? Train when most alert |
| Session structure | Warm-up (easy known command) → new |
| | material → cool-down (easy command) |
+--------------------+------------------------------------------+
The 80/20 Rule:
- 80% of reps should succeed (dog is getting it right)
- If success rate drops below 80%, the criteria is too high — go easier
- 20% challenge keeps the dog engaged without frustrating
得: 短而成之課,犬欲更多。
敗: 犬離(嗅、視他、臥)→課過長、過難或獎不足。結課而再評。
四:擾下穩命
靜中穩後漸增擾。
Distraction Ladder (work through sequentially):
1. Quiet room, no distractions (starting point)
2. Room with a family member present
3. Backyard or garden
4. Front yard with street noise
5. Quiet park or trail
6. Busy park with other dogs at a distance
7. Busy park with other dogs nearby
8. Novel environments (pet store, cafe patio)
At each new level:
- Expect performance to decrease — this is normal
- Increase reward rate back to every correct rep
- Do not add more distraction until the current level is reliable
- If the dog fails 3 reps in a row, you moved up too fast — go back one level
得: 命於漸擾環境中穩行。
敗: 某擾(他犬、松鼠)常破訓→需別對條反訓(見 behavioral-modification)。
驗
- 標已充而犬穩應
- 五命於寡擾境皆行
- 課 5-10 分,終於成
- 每命成率八成以上
- 命依擾梯泛化
- 主人時機(標於一秒內)穩
忌
- 復命:言「坐、坐、SIT」→教犬首「坐」可略。言一次而待
- 晚獎:獎當於標後 2-3 秒內。晚獎破聯
- 久誘:手帶獎之動(誘)宜於 10-20 次內淡去。否則犬唯見食而應
- 罰失召:呼「來」而責(因緩、因口銜物)→永毒召令
- 久訓:疲犬無學。勝而止
- 命不一:諸家員當用同辭與勢於每命
參
behavioral-modification— 處擾基順之不欲行
GitHub Repository
Related Skills
llamaguard
OtherLlamaGuard is Meta's 7-8B parameter model for moderating LLM inputs and outputs across six safety categories like violence and hate speech. It offers 94-95% accuracy and can be deployed using vLLM, Hugging Face, or Amazon SageMaker. Use this skill to easily integrate content filtering and safety guardrails into your AI applications.
cost-optimization
OtherThis Claude Skill helps developers optimize cloud costs through resource rightsizing, tagging strategies, and spending analysis. It provides a framework for reducing cloud expenses and implementing cost governance across AWS, Azure, and GCP. Use it when you need to analyze infrastructure costs, right-size resources, or meet budget constraints.
quantizing-models-bitsandbytes
OtherThis skill quantizes LLMs to 8-bit or 4-bit precision using bitsandbytes, achieving 50-75% memory reduction with minimal accuracy loss. It's ideal for running larger models on limited GPU memory or accelerating inference, supporting formats like INT8, NF4, and FP4. The skill integrates with HuggingFace Transformers and enables QLoRA training and 8-bit optimizers.
dispatching-parallel-agents
OtherThis Claude Skill dispatches multiple agents to investigate and fix 3+ independent problems concurrently. It is designed for scenarios involving unrelated failures that can be resolved without shared state or dependencies. The core capability is parallel problem-solving, assigning one agent per independent problem domain to maximize efficiency.
