with one click
Delegate coding to Claude Code CLI (features, PRs).
npx skills add https://github.com/hiyenwong/ai_collection --skill claude-codeCopy and paste this command into Claude Code to install the skill
Delegate coding to Claude Code CLI (features, PRs).
npx skills add https://github.com/hiyenwong/ai_collection --skill claude-codeCopy and paste this command into Claude Code to install the skill
| name | claude-code |
| description | Delegate coding to Claude Code CLI (features, PRs). |
| version | 2.3.0 |
| author | Hermes Agent + Teknium |
| license | MIT |
| platforms | ["linux","macos","windows"] |
| metadata | {"hermes":{"tags":["Coding-Agent","Claude","Anthropic","Code-Review","Refactoring","PTY","Automation"],"related_skills":["codex","hermes-agent","opencode"]}} |
Delegate coding tasks to Claude Code (Anthropic's autonomous coding agent CLI) via the Hermes terminal.
npm install -g @anthropic-ai/claude-codeclaude once to log in, or set ANTHROPIC_API_KEYclaude auth status --textclaude -p "echo ok" --max-turns 1 --allowedTools Bash| Use Claude Code | Don't Use (fall back to direct) |
|---|---|
| Multi-step refactoring across files | Single file edit |
| Complex feature implementation | Quick fix / one-off change |
| Tasks requiring complex tool orchestration | Claude Code fails auth/network check |
| OpenSpec-driven feature development | API 403 / org restriction blocks CLI |
-p) — Non-Interactive (PREFERRED)claude -p 'Add error handling to all API calls in src/' --allowedTools 'Read,Edit' --max-turns 10
Skips all interactive dialogs. Ideal for automation.
tmux new-session -d -s claude-work -x 140 -y 40
tmux send-keys -t claude-work 'cd /project && claude' Enter
# Handle trust dialog
sleep 5 && tmux send-keys -t claude-work Enter
# Send task
tmux send-keys -t claude-work 'Implement auth module' Enter
# Monitor
tmux capture-pane -t claude-work -p -S -50
Symptom: claude auth status --text shows logged in, but claude -p "echo ok" returns "Failed to authenticate. API Error: 403 Request not allowed".
Cause: Organization-level restrictions on the Claude Pro plan may block CLI API calls even though OAuth is valid. This is NOT a local configuration issue.
Detection:
claude auth status --text # Looks fine
claude doctor # Reports healthy
claude -p "echo ok" --max-turns 1 --allowedTools Bash
# Returns 403 → Claude Code is unavailable
Fix:
claude auth login --console or set ANTHROPIC_API_KEYWhen Claude Code is unavailable but the task requires OpenSpec-driven development with quality gates:
write_file() per filecargo fmt && cargo clippy -- -D warnings && cargo testgit add -A && git commit -m "feat: ..." with conventional messages| Aspect | Claude Code | Fallback (Direct) |
|---|---|---|
| File creation | Delegated | write_file() |
| Test writing | Included in task | Write alongside implementation |
| Iteration | Self-corrects | Manual patch per clippy error |
| Speed | Parallel | Sequential per-module |
| Quality gates | Runs at end | After each module |
cargo fmt && cargo clippy -- -D warnings && cargo test
feat:, fix:, test:, docs:, refactor:For autonomous ML experiment iteration (run → analyze → fix → re-run → scale → paper update), see references/autonomous-ml-research-loop.md.
Key steps:
tmux new-session -d -s <name> && tmux send-keys ... claude --dangerously-skip-permissionstmux capture-pane -t <session> -p -S -20--dangerously-skip-permissions defaults to "No, exit" — must send Down then Enter--max-turns is print-mode only--json-schema needs enough --max-turns — Claude must read files first/compactclaude auth status lies. Always quick-test before starting long sessions.Attachment representations in early childhood as independent endogenous driver of interbrain synchrony during remote cooperation. Novel Remote Partner-Belief Manipulation paradigm isolates attachment representations by manipulating partner-belief. EEG synchrony concentrated at P4 channel (right TPJ). Activation: attachment, interbrain synchrony, EEG hyperscanning, child-adult interaction, attachment representations, social neuroscience, partner-belief manipulation, early childhood, mother-child interaction, brain synchronization, attachment security, social-emotional development.
SHARP (Sleep-based Hierarchical Accelerated Replay) 方法论 — 睡眠启发的分层加速回放框架用于长程非平稳时序模式识别。受啮齿动物慢波睡眠中加速回放启发, 通过分离记忆模块和模式识别模块实现无反向传播的长程信用分配。适用于流式时序学习、 长程依赖建模、神经科学启发的 AI 架构。触发词:睡眠回放、加速回放、SHARP、 时序学习、长程依赖、流式学习、慢波睡眠、hierarchical replay
Inverse-engineering methodology for piston operations in trapped-ion quantum devices. One ion serves as classical piston driven by Coulomb interaction with quantum-controlled ion. Stationary state determined self-consistently. Inverse-engineering protocols enable precise control of classical ion motion. Provides route toward controlled piston dynamics in microscopic quantum devices.
Quantum fault tree analysis methodology using quantum computing. Extends classical reliability engineering fault trees to quantum domain. Identifies minimal cut sets in system reliability analysis using quantum algorithms. Applicable to safety-critical systems, cyber-physical systems, and quantum system reliability engineering.
Adaptive Hybrid Quantum-Classical Feature Fusion methodology for medical image classification. Addresses optimization asymmetries between quantum and classical paradigms using Temperature-Scaled Hybrid Fusion (TSHF), Dynamic Hybrid Fusion (DHF), and Static Hybrid Fusion (SHF) strategies. Use when designing hybrid quantum-classical ML pipelines for healthcare/medical imaging, especially when combining ResNet backbones with variational quantum circuits for diagnostic tasks.
Adaptive Spiking Neuron (ASN) methodology for vision and language modeling. Implements trainable membrane potential dynamics with adaptive firing mechanisms for efficient Spiking Neural Networks (SNNs). Activation: adaptive spiking neuron, ASN, spiking neural network vision language, SNN adaptive neuron, neuromorphic vision language model.