菜单
End-of-conversation review to capture learnings. Use when the user runs /reflect to review what happened and suggest improvements to CLAUDE.md, skills, or new skills.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Guidelines for writing Supabase database migrations, functions, RLS policies, and SQL. Use when creating or modifying database schemas, writing migrations, creating Postgres functions, or setting up Row Level Security policies.
Documentation for Agicash wallet transaction flows and entities. Use when touching app/features/send/, app/features/receive/, app/features/wallet/, token receive routes, Lightning Address routes, account selection logic, or working with Quote/Swap/Transaction entities and payment state machines.
TanStack Query v5 patterns specific to this project. Covers our cache class pattern, repository integration, suspense queries, event-driven updates, version-aware caching, and mutation patterns. Use for all server state management and cache operations.
Use this skill when the task requires understanding *why* the app does something (blind signatures, keyset rotation, spending conditions), not just *what* it does. References Cashu NUT specifications for the ecash protocol on Bitcoin. Complements /agicash-wallet-documentation — that skill describes our implementation, this describes the protocol it implements. Invoke for learning, explaining, implementing, or debugging Cashu concepts like minting, swapping, melting, blind signatures, keysets, or quote flows.
Validation tool for Lightning Address endpoints. Use after modifying Lightning Address routes (app/routes/[.]well-known.*, app/routes/api.lnurlp.*) or lightning-address-service.ts to verify LUD-16, LUD-06, and LUD-21 endpoints still work.
Tailwind CSS design system for this mobile-first wallet app. Covers theme system (USD/BTC/dark), CSS variables, CVA components, shadcn/ui patterns, custom animations, and layout conventions. Use when creating or modifying UI components.
| name | reflect |
| description | End-of-conversation review to capture learnings. Use when the user runs /reflect to review what happened and suggest improvements to CLAUDE.md, skills, or new skills. |
| disable-model-invocation | true |
Review the current conversation and suggest high-ROI improvements. Keep suggestions brief and actionable.
CLAUDE.md updates — Did you make a mistake that better guidance would prevent? Did the user correct you on a pattern or preference? Did you learn something about the codebase that would save time next session? Did anything in the conversation contradict or conflict with existing CLAUDE.md guidance (stale rules, outdated patterns, wrong assumptions)?
Wallet documentation sync — Did the conversation change any behavior documented in the agicash-wallet-documentation skill? Read its SKILL.md to understand what it covers, then selectively check only the reference files relevant to the changes made. Suggest edits if anything is now out of date.
Skill improvements — Did you use a skill that gave bad advice, was missing context, or had an awkward workflow? Would small edits make it more effective?
New skills — Did the conversation involve a repeatable workflow that would benefit from a skill? Only suggest if it came up naturally and seems like it would be used again.
cashu-lightning-send.md — don't read spark-operations.md)/update-context, /skill-creator, or edit the skill/documentation directly)