Development guidelines for DotCraft project. Use this skill when developing DotCraft core features, adding new modules (including external channel adapters via AppServer/JRPC), modifying existing code, or writing documentation. Covers C# code style, tool naming (PascalCase for AI functions), module development norms (via spec), external channel extension with Python SDK, spec-first workflow, testing requirements, and bilingual documentation.
>-
Add or extend API Platform resources in php-service-template using DDD and CQRS-friendly patterns.
- 📁 docs/
- 📁 references/
- 📄 .gitignore
- 📄 README.md
- 📄 SKILL.md
Use whenever a user shares or describes a 104 履歷 (resume) or LinkedIn profile and wants it reviewed, scored, or improved — even if they only paste a fragment or describe it verbally.
Use token-efficient CLI patterns instead of verbose MCP output when direct shell tools are enough. Provides JSON or compact-output conventions for gh, mgrep, psql, and similar tools.
Compose an arrangement for a tape — plan which elements enter and exit over time, then generate a playable CCArrangement.
Use when setting up a project environment — installing dependencies, verifying builds, detecting the tech stack. Covers Phase 0 of a new session.
Perform a self-review of a PR before requesting human review. TRIGGER when user invokes /pr-selfcheck or when the git workflow reaches the self-review step after PR creation. Accepts a PR number as an argument.
Build multi-step LLM reasoning chains in n8n using Groq, OpenAI, or Claude for structured data extraction, categorization, scoring, and analysis. Use this skill whenever the user wants to chain multiple LLM calls together in an n8n workflow — phrases like "extract entities then categorize", "multi-step LLM prompt", "chain_llm", "LLM pipeline", "classify and score", "entity extraction then enrichment". Also use when processing call transcripts, customer messages, or any unstructured text through multiple analysis passes in n8n. Prefer this pattern over single-shot prompts whenever the output requires both extraction AND reasoning, since single-shot hallucinates categories while chains let each step verify the previous.
Run Hermes Local Memory peer review, reflection, deterministic maintenance, candidate review, card review, and verification through provider tools.
Writes Python code following FiftyOne's official conventions. Use when contributing to FiftyOne, developing plugins, or writing code that integrates with FiftyOne's codebase.
Use when an agent needs the smallest useful eval pack to catch regressions or false confidence early.
