A personal library of reusable AI agent skills for Shane Myrick. Skills encode repeatable workflows — each one tells an AI agent exactly how to handle a specific task, using the right tools, in the right order, with the right context.
A skill is a Markdown file (SKILL.md) that lives in a named folder under skills/. When loaded by an AI agent (e.g., Claude in Cowork), it provides:
- A trigger description — when should the agent use this skill?
- Step-by-step workflow — exactly what to do, in what order
- Tool references — which MCPs, APIs, or integrations are required
- Optional org- or project-specific context — when a workflow needs them: conventions, cloud IDs, project keys, personas
Skills are agent-readable instructions, not code. They make common work reproducible without starting from scratch every time.
Install every skill from this repo into your coding agents (Cursor, Claude Code, Codex, OpenCode, and others supported by the skills CLI):
npx skills add smyrick/skillsOr with the full URL:
npx skills add https://github.com/smyrick/skillsList skills without installing: npx skills add smyrick/skills --list. The deprecated npx add-skill forwards to npx skills add.
From the repository root, check that every skills/*/SKILL.md has valid frontmatter (name, description, compatibility):
npm install
npm run validateFor stricter checks (recommended before opening a PR), including optional metadata:
npm run validate:strictCI runs validate:strict on every push and pull request.
| Skill | Description | Key Tools |
|---|---|---|
| codebase-summary | Analyze and document codebase architecture with ARCHITECTURE.md files, including entry points, APIs, core modules, and Mermaid diagrams | File system tools (Read, Glob, Grep) |
| create-a-skill | Scaffold and author a new skill in this library (frontmatter, workflow, README index, org-specific context when needed) | Read, Write, StrReplace; optional Atlassian MCP |
| humanize-text | Rewrite text to remove common AI-output tells (em-dashes, "delve", colon-led impact statements, etc.) while preserving meaning | Read, Write, Edit |
| plan-mode | Create structured implementation plans with acceptance criteria that can be handed off to a less capable agent | AskQuestion tool |
| write-a-prd | Create a PRD through interview, codebase exploration, and module design, then submit as a GitHub issue | Read, Glob, Grep; GitHub |
Skills in this repo are installed at ~/.skills/skills/ on your machine. Once installed, Claude will automatically load the right skill based on your request — no manual steps needed.
If you want to point Claude to a skill by name:
"Use the plan-mode skill to plan this feature before we implement it"
Any agent with access to this repo can read a SKILL.md directly and follow its instructions. The files are written to be self-contained — a capable agent should be able to execute the full workflow after reading one.
README.md ← You are here
CONTRIBUTING.md ← How to add or improve skills
package.json ← npm run add-skill → scaffolds under skills/
scripts/validate-skills.mjs ← npm run validate / validate:strict
skills/
<skill-name>/
SKILL.md ← The skill itself (agent-readable instructions)
create-a-skill/
SKILL.md ← Skill that guides authoring new skills
Skills are organized as one folder per skill. The folder name is the skill's ID — use lowercase, hyphenated names (e.g., customer-qbr-prep, slack-deal-summary).
Skills I intend to build next (roughly in priority order):
- customer-qbr-prep — Pull deal context from Salesforce + Jira + Slack, generate a QBR briefing doc
- account-health-digest — Weekly digest of at-risk accounts with Slack + Jira signals
- sa-handoff-writer — Turn a Slack thread into a structured SA handoff note in Confluence
- meeting-brief-generator — Pre-meeting brief from calendar invite + Salesforce + recent Slack
- solutions-weekly-standup — Aggregate team Jira updates into a manager-ready standup summary
Want to contribute one? See CONTRIBUTING.md.
Generic pointers sometimes referenced from skills (no tenant or employer-specific IDs):
| Constant | Value |
|---|---|
| Preferred Claude model | claude-sonnet-4-20250514 |
| Atlassian MCP URL | https://mcp.atlassian.com/v1/mcp |
MIT — see LICENSE.