skill
orchestrator-operational-agent-skills
25 portable Agent Skills that turn an AI assistant into an operating team: SDLC, PRDs, code review, deploy, finance, marketing, content, HR. Works with Claude Code and any SKILL.md runtime.
Set up
Skills are installed by placing their folder where your agent loads skills.
The download comes from github.com/me-workspace/orchestrator-operational-agent-skills itself, not from our server, so what you get is exactly what its author published.
Install instructions come from the README
README
synced Oct 26, 2026Agent Skills for Business and Engineering
25 portable skills that turn a generic AI assistant into an operating team: ship software end to end and run the business around it, from PRDs and code review to finance, marketing, and HR.
A Skill is a folder with a SKILL.md file: a focused, reusable procedure an AI agent loads only when it is relevant. This repo is a curated set of them, written to Anthropic's Agent Skills open standard, so they drop into Claude Code, the Claude Agent SDK, or any agent runtime that reads SKILL.md.
It is more than a pile of prompts. Every skill is a checklist, rubric, or decision tree with quality gates, and AGENTS.md shows how to compose them into an 8-agent company (COO plus seven specialists) without the skills fighting each other for context.
Why this pack
- Procedures, not prose. Skills produce consistent, reviewable output instead of freestyle text.
- Trigger-rich descriptions. Each
SKILL.mdsays exactly when to fire, and when NOT to, which cuts wrong-skill activation when you install many. - Context-cheap. Only the name and description sit in context; the body loads on demand. Install a dozen skills without drowning the window.
- Honest by construction. Skills that touch tax, law, or money carry a draft-until-verified rule inside the skill itself.
- Portable. If your agent can read a
SKILL.md, it can use these.
Quick start (Claude Code)
git clone https://github.com/me-workspace/orchestrator-operational-agent-skills.git
cd orchestrator-operational-agent-skills
# One skill, project-scoped
cp -r marketing-growth /path/to/your-project/.claude/skills/
# Or install a role loadout, globally
cp -r prd-creator project-timeline-creator strategic-approach admin-ops ~/.claude/skills/
The next time a matching request comes in ("write a campaign brief", "draft a PRD", "review this diff"), the agent loads the skill automatically.
Do not install all 25 into one agent. Metadata for every installed skill stays in context. Pick a loadout per role. See
AGENTS.mdfor a worked design.
The skills
Software and AI (full SDLC)
| Skill | What it does |
|---|---|
sdlc-master | End-to-end lifecycle: phases and gates, tech design doc, Definition of Done, branching, testing strategy |
principal-engineering | Principal-level judgment: principles, trade-offs, decision framework, standard calls |
strategic-approach | Choosing an approach: framing, three diverse options, recommendation plus kill criteria, MVP scoping |
prd-creator | Full PRD: measurable goals, mandatory non-goals, Given/When/Then acceptance criteria |
project-timeline-creator | Timelines: breakdown to 2 days or less, dependencies, 1.5x buffer, critical path, slip protocol |
project-audit | Project audit: 7 scored dimensions, evidence-based method, prioritized report and verdict |
code-review-hardening | Code review: correctness and security, verify before reporting, severity rubric |
secure-deploy-ops | Deploy: pre-deploy backup, live vs staged diff, smoke test, hardening baseline |
incident-debugging | Production debugging: evidence-first, discriminating hypotheses, minimal fix, write-up |
uiux-engineer | UI/UX engineering: five required states, forms, responsive, accessibility, usability review |
product-designer | Designer's eye: hierarchy, typography, color, spacing system, design system, critique |
agent-skill-author | Meta-skill: write and audit other agents' skills against a 12-point quality bar |
tech-scout | Tech radar: source sweep, 6-dimension rubric, hype immunity, CVE and EOL watch |
Operations
| Skill | What it does |
|---|---|
finance-analyst | SaaS metrics, runway, unit economics, pricing, monthly snapshot, red flags |
accounting-core | Double-entry journals, chart of accounts, reconciliation, assets and depreciation, tax draft, closing |
sales-pipeline | Scored lead qualification, discovery, proposal with scope exclusions, cadence, negotiation |
marketing-growth | Campaign briefs, copywriting rules, SEO checklist, growth experiments, funnel review |
social-media-specialist | Per-platform grammar, content calendar and pillars, repurposing chain, metrics review |
content-creator | Production: hook patterns, per-format structure (article, script, newsletter, thread), storytelling |
content-editor | Four-pass quality gate: structure, clarity, style, correctness; verdict plus kill criteria |
admin-ops | SOPs, two-tier meeting minutes, email triage, document conventions |
hr-officer | Rubric-based recruitment, 30-day onboarding, performance, sensitive cases, compensation |
hr-admin | Payroll and statutory deductions (draft-until-verified), leave and attendance, contracts, compliance calendar |
Cross-agent
| Skill | What it does |
|---|---|
coo-orchestrator | Leader agent: routing, cross-agent chain shepherding, daily ops pulse, weekly review, escalation, accountability register |
continuous-learning | Baseline for every agent: fast-decay fact classification with mandatory verification, learning from corrections, scheduled knowledge review, honesty rules |
Tax and compliance skills (
accounting-core,hr-admin,hr-officer) include Indonesia-oriented specifics and are marked draft-until-verified against the current year's regulations or a licensed professional. Adapt them to your jurisdiction.
Compose them into a team
AGENTS.md is a full worked design: which skills go on which agent, why the roles are split the way they are (an auditor should not audit its own work; finance data stays isolated from public-facing agents), and how to run it day to day. It also gives a lean 5-agent option. (Indonesian version: AGENTS.id.md.)
How a skill works
skill-name/
└── SKILL.md # front matter (name + description) + the procedure
---
name: marketing-growth
description: Use this skill for marketing work, campaigns, SEO, ads, copywriting,
landing pages, funnels. Do NOT use for one-to-one sales (use sales-pipeline).
---
# Marketing and Growth
No deliverable without: target audience, single message, measurable metric...
The agent keeps every skill's description in context and reads the body only when a task matches. That is what makes a large pack practical.
Contributing
New skills, sharper triggers, and better methodology are welcome. See CONTRIBUTING.md. Keep skills procedural, keep descriptions trigger-rich.
License
MIT (c) 2026 Wira Darma. Use them, fork them, ship them.
Indonesian catalog: README.id.md.