Software Development Skills

Expert guide for setting up and improving a Claude personal assistant — covering CLAUDE.md, skills, memory, scheduled tasks, token efficiency, self-improving tasks, MCP servers, git integration, dev workflow, personal data projects, and LLM wikis. Use this skill whenever a user wants to: set up Claude as a personal assistant, write or improve a CLAUDE.md, create or refine a skill, set up memory or profiles, build or optimize a scheduled task, connect Claude to personal data (investments, finances, health, etc.), set up a new data project, add patterns to an existing project, configure MCP servers, integrate git, or asks anything like "how do I set up Claude to...", "help me improve my...", "create a skill for...", "set up a task that...", "Claude keeps forgetting...", or "how do I get Claude to work with my data". Trigger broadly — when in doubt, load this skill.

Analyze and optimize a Cowork task or project to make it run faster, use fewer tokens, reduce unnecessary steps, and improve overall structure. Use this skill whenever the user shares a Cowork task or project and asks to optimize, improve, speed up, refactor, or reduce Claude usage. Also trigger when the user says things like "make this more efficient", "this task is slow", "can we trim this down", "review my task", or pastes a Cowork task/project definition and asks for feedback.

Full-lifecycle git and GitHub management for the user's repositories — it diagnoses the repo's real state, then commits, branches, syncs, merges, and cleans up itself where it can, and hands over exact validated copy-paste commands where the session can't act (credentials, network, or environment limits). Use this skill whenever the user mentions git, GitHub, commit, push, pull, branch, merge, PR, squash, rebase, tag, release, .gitignore, or repository — AND for vague phrasings that imply version control: "save my work", "back this up", "is this synced?", "get this onto GitHub", "clean up this repo", "did my changes land?", "undo that change". Also use it proactively: when a session did file work inside a git repo and is wrapping up, when a repo shows uncommitted changes or lock files, or when the user is about to work on files that may be stale against the remote.

Company policy guardrail for every Claude response. Use this skill ALWAYS before answering any question, drafting any output, running any task, or recommending any action that could touch company policy. Triggers include: drafting external or internal communications, handling data of any classification, making decisions about tools or vendors, summarising or forwarding content, content generation of any kind, recommending technical approaches with security implications, and any task or prompt that processes organisational data. If the user asks anything that could implicate the AI use policy, Claude guidelines, Code of Conduct, data classification, acceptable use, or related policies, this skill MUST be consulted first. Applies three enforcement tiers: hard block (T1), required check (T2), soft guidance (T3).

Use when the user wants to create a reusable template from an existing Claude artifact or setup — including chat system prompts, Cowork tasks, Cowork projects, or skills. Trigger when the user says things like "turn this into a template", "create a template I can share", "make this reusable", "export this as a template", "I want to share this setup with someone", or "create a template for [task/project/skill/chat]". Also trigger when a conversation contains a full Cowork task, project, or skill definition and the user seems to want to preserve or share it. Always use this skill — even if the user just says "template" — when the context involves chat prompts, Cowork configurations, or skill definitions.