Skills - OrgX Docs

Skills are reusable instruction sets that extend what Claude (and other MCP clients) can do with OrgX. Each skill pack gives your AI client a complete workflow — from context gathering to artifact output — for a specific OrgX task.

Skills live in the useorgx/skills public repository. The wizard CLI installs them locally with one command:

npx @useorgx/wizard@latest skills add

In the OrgX app and cloud runtime, skills/ is a generated, pinned cache of useorgx/skills, not a second authoring surface. Runtime reads stay local and network-free; CI uses pnpm skills:check-source to prove the cache still matches useorgx/skills@<sha>. Use SKILL_PACKS_DIR only for explicit local overrides.

How Skills Work

When you install an OrgX skill pack, the wizard writes a SKILL.md file into your local Claude configuration at .claude/skills/orgx/. Claude picks this up as context for any conversation in that directory, giving it:

The right sequence of MCP tool calls to call
Domain-specific quality rules and artifact contracts
Evidence proof requirements (durable links, not OrgX wrapper pages)
Trigger patterns so skills activate on natural language cues

For Cursor users, the wizard also writes .cursor/rules/orgx.md with equivalent rules.

Source of Truth and Runtime Cache

Canonical authoring happens in useorgx/skills. The OrgX app vendors a generated cache at skills/ so cloud agents can load skill instructions without fetching GitHub during an agent run.

pnpm skills:sync-source refreshes the cache from useorgx/skills and updates skills/.source.lock.json.
pnpm skills:check-source fails when any managed skill file drifts from the pinned source ref.
App-only skills remain explicit local overlays in skills/.source.lock.json.
SKILL_PACKS_DIR=/path/to/skills is still supported for deliberate local development overrides.

The portable skill format is:

skill-id/
  SKILL.md       # required cross-runtime instructions
  skill.toml     # required OrgX runtime metadata
  reference/     # optional, loaded on demand
  examples/      # optional, loaded on demand
  scripts/       # optional validators/helpers
  evaluations/   # optional CI/eval fixtures

Installing Skills

Default Starter Pack

npx @useorgx/wizard@latest skills add

Installs the four default packs: morning-briefing, initiative-kickoff, bulk-create, nightly-recap.

Specific Packs

npx @useorgx/wizard@latest skills add engineering-agent orchestrator-agent

All Packs

npx @useorgx/wizard@latest skills add morning-briefing nightly-recap \
  initiative-kickoff initiative-protocol workstream-protocol milestone-protocol task-protocol \
  bulk-create engineering-agent product-agent marketing-agent sales-agent \
  design-agent design-audit operations-agent orchestrator-agent

Available Skills

Daily Operations

morning-briefing

Trigger: “morning briefing”, “daily status”, “what’s pending”, “what should I work on first?”Generates a concise daily report with value signals, pending decisions (critical first), blocked tasks grouped by initiative, and agent activity. Calls orgx_recommend + orgx_search for decisions and blocked tasks.

nightly-recap

Trigger: “nightly recap”, “daily summary”, “what happened today”, “end of day report”Summarizes the day’s OrgX activity — completed tasks, decisions made, agent work, and any errors that need attention. Good for async standup notes or end-of-day logs.

Initiative Management

initiative-kickoff

Trigger: “kickoff”, “new initiative”, “start project”, “create initiative for”, “set up [goal]”Transforms a one-line goal into a full OrgX initiative: parses the objective, checks for duplicates, creates the initiative with milestones and workstreams, assigns agent domains, and optionally launches immediately.Not for: updating existing initiatives, simple task creation.

initiative-protocol

Trigger: managing an active OrgX initiative lifecycleHandles creation, decomposition into workstreams and streams (with DAG dependencies), launch, monitoring, and completion. Reference protocol for any skill that needs to touch initiative state.

workstream-protocol

Trigger: workstream creation, management, stream coordinationCovers workstream creation, stream setup with agent domain and auto_continue, dependency wiring, and health checks. Used internally by initiative-protocol and orchestrator-agent.

milestone-protocol

Trigger: milestone creation, scheduling, gate checkingManages milestone definitions with due dates and completion gates. Coordinates with workstreams to track progress toward milestones.

task-protocol

Trigger: individual task execution within a workstreamHandles task lifecycle: start → execute → verify → complete. Emits telemetry at each phase, follows domain-specific workflows, and submits learnings on completion.

Domain Agent Skills

These skills give your MCP client the same specialized expertise as OrgX’s built-in agents. Each produces evidence-based artifacts with durable proof links.

engineering-agent

Domain: Technical decisions, implementation risk, code qualityArtifacts produced: RFCs, ADRs, code review summaries, incident postmortems, tech debt inventories, capacity plans, runbooks, migration playbooks, dependency audits, performance budgets, architecture docs (any C4 level), threat models, SLO definitions, build-vs-buy analysesProof rules: GitHub PR/commit/blob permalinks, public docs URLs, or absolute file paths only. No OrgX wrapper pages.

# Example triggers
"write an RFC for migrating our auth system"
"review this PR for security issues"
"create a runbook for database failover"

product-agent

Domain: Problem framing, user value, prioritization, measurable outcomesArtifacts produced: PRDs, initiative plans, product canvases, user research briefs, competitive analyses, feature prioritization matrices, pivot evaluations, metric dashboard specs, launch readiness checklists

# Example triggers
"write a PRD for the new onboarding flow"
"prioritize these features for Q2"
"competitive analysis vs. [competitor]"

marketing-agent

Domain: Campaign planning, content creation, growth, SEO/AEOArtifacts produced: Campaign briefs, landing page copy, blog posts, email sequences, social content calendars, SEO/AEO keyword plans, go-to-market playbooks, press releases, brand messaging guides

# Example triggers
"create a campaign brief for our product launch"
"write a blog post about [topic]"
"draft an email nurture sequence for trial users"

sales-agent

Domain: Pipeline, outreach, objection handling, competitive positioningArtifacts produced: Outreach sequences, qualification frameworks, competitive battlecards, proposal templates, win/loss analyses, ICP definitions, discovery call scripts

# Example triggers
"create an outreach sequence for enterprise prospects"
"build a battlecard for [competitor]"
"write a discovery call script"

design-agent

Domain: UX rationale, design decision support, UI briefsArtifacts produced: UX rationale documents, design briefs, component specs, accessibility assessments, user flow diagrams, design system contribution guidelines

# Example triggers
"write a UX rationale for the new checkout flow"
"create a design brief for the mobile navigation"

operations-agent

Domain: Process operations, incident management, runbooksArtifacts produced: Incident reports, process playbooks, runbooks, SOC procedures, vendor evaluations, operational risk assessments, team capacity reports

# Example triggers
"create a runbook for on-call escalation"
"write an incident report for last night's outage"
"evaluate [vendor] for our data pipeline"

orchestrator-agent

Domain: Cross-domain execution coordinationArtifacts produced: Initiative plans, delegation messages, synthesis reports, retrospectives, dependency audits, resource allocations, risk registers, stakeholder updates, program status reportsThe orchestrator never does domain work itself — it coordinates, sequences, unblocks, and synthesizes across other agents.

# Example triggers
"coordinate the Q2 product and marketing launch"
"create a cross-team initiative for [goal]"
"synthesize this week's agent outputs into a status report"

Specialized Skills

design-audit

Trigger: /design-audit <target_path_or_url> [reference_urls...]Applies cognitive psychology, Gestalt theory, and principles from Rams, Tufte, Norman, and others to audit any UI surface. Outputs a scored assessment with prioritized improvements. Minimum target: 9/10 per pass.

bulk-create

Trigger: “bulk create”, “create tasks from list”, “import checklist”, “batch create”Parses markdown checklists, bullet lists, or numbered lists and creates multiple OrgX entities in batch. Detects priorities from keywords and wires up dependencies for nested items.

Proof Rules (All Skills)

Every domain skill enforces these artifact proof rules:

Durable sources only: GitHub PR/commit/blob permalinks, public URLs, or absolute file paths (/path/to/file or file://...)
No OrgX wrapper pages: /live/..., /artifacts/..., and /console/... are dashboards, not evidence
preview_markdown is supporting context: it does not replace a durable proof URL

These rules exist so artifacts remain verifiable outside OrgX itself.

Writing Custom Skills

Skills are plain Markdown files with a YAML frontmatter header. Place them in .claude/skills/ in your project directory:

---
name: my-custom-skill
description: |
  What this skill does and when to use it.
  Triggers on: "specific phrases the user might type"
---

# My Custom Skill

## Workflow

1. Gather context with `mcp__orgx__orgx_search`
2. ...

See the useorgx/skills repo for examples.

Next Steps

Wizard CLI

Full CLI reference for installing and managing skills.

Agent Quickstart

Run your first agent task with skills wired in.

Agent Recipes

Pre-built workflows combining multiple skills.

MCP Tools

Full reference for the tools skills call under the hood.

​How Skills Work

​Source of Truth and Runtime Cache

​Installing Skills

​Default Starter Pack

​Specific Packs

​All Packs

​Available Skills

​Daily Operations

morning-briefing

nightly-recap

​Initiative Management

initiative-kickoff

initiative-protocol

workstream-protocol

milestone-protocol

task-protocol

​Domain Agent Skills

​Specialized Skills

design-audit

bulk-create

​Proof Rules (All Skills)

​Writing Custom Skills

​Next Steps

Wizard CLI

Agent Quickstart

Agent Recipes

MCP Tools

How Skills Work

Source of Truth and Runtime Cache

Installing Skills

Default Starter Pack

Specific Packs

All Packs

Available Skills

Daily Operations

Initiative Management

Domain Agent Skills

Specialized Skills

Proof Rules (All Skills)

Writing Custom Skills

Next Steps