Agent Skill
2/7/2026

create-skill

Create new Standard Operating Procedure documents following project standards. Use when documenting new processes, workflows, or guidelines. Invoked by: "new process", "create sop", "document process", "new workflow", "create documentation", "new sop", "create skill".

E
eleva
0GitHub Stars
1Views
npx skills add eleva-labs/agentic-coding-templates

SKILL.md

Namecreate-skill
DescriptionCreate new Standard Operating Procedure documents following project standards. Use when documenting new processes, workflows, or guidelines. Invoked by: "new process", "create sop", "document process", "new workflow", "create documentation", "new sop", "create skill".

name: create-skill description: >- Create new Standard Operating Procedure documents following project standards. Use when documenting new processes, workflows, or guidelines. Invoked by: "new process", "create sop", "document process", "new workflow", "create documentation", "new sop", "create skill".

SOP Creation SOP

Version: 1.0.0 Last Updated: 2026-01-11 Status: Active


Overview

Purpose

This guide provides a standardized approach for creating new Standard Operating Procedure (SOP) documents. It ensures consistency across all process documentation, making SOPs easy to follow, maintain, and adapt into Claude Code skills.

When to Use

ALWAYS: Creating new processes, documenting existing workflows, establishing guidelines, creating templates, defining procedures SKIP: Quick notes, temporary documentation, one-off instructions


YAML Frontmatter Reference

Every skill file (SKILL.md) must begin with YAML frontmatter. This section documents all available fields based on official Claude Code documentation.

Available Fields

FieldRequiredDescription
nameNoDisplay name for the skill. Defaults to directory name. Lowercase letters, numbers, and hyphens only (max 64 characters).
descriptionRecommendedWhat the skill does and when to use it. Claude uses this to decide when to apply the skill.
argument-hintNoHint shown during autocomplete. Example: [issue-number] or [filename] [format].
disable-model-invocationNoSet to true to prevent Claude from automatically loading this skill. Default: false.
user-invocableNoSet to false to hide from the / menu. Default: true.
allowed-toolsNoTools Claude can use without asking permission when this skill is active.
modelNoModel to use when this skill is active.
contextNoSet to fork to run in a forked subagent context. Only use for isolated tasks.
agentNoWhich subagent type to use when context: fork is set. See Agent Types below.
hooksNoHooks scoped to this skill's lifecycle.

Agent Types (for context: fork)

When using context: fork, the agent field specifies which agent runs the skill:

Built-in Agents:

AgentDescription
general-purposeDefault agent with full tool access. Used if agent is omitted.
ExploreRead-only agent optimized for codebase exploration (Glob, Grep, Read).
PlanPlanning agent for designing implementation approaches.

Custom Agents: Reference any agent defined in .claude/agents/ by path:

agent: mobile/engineer    # Uses .claude/agents/mobile/engineer/AGENT.md
agent: backend/architect  # Uses .claude/agents/backend/architect/AGENT.md

Important Notes

  • Only description is recommended; all other fields are optional
  • context: fork runs the skill in isolation WITHOUT conversation context - use sparingly and only for truly isolated tasks
  • agent only matters when context: fork is set - it has no effect on inline skills
  • Most skills should NOT have context or agent - they should run inline with full conversation context

Example: Inline Skill (Most Common)

Most skills should run inline, maintaining conversation context:

---
name: my-skill
description: >-
  What this skill does.
  Invoked by: "trigger phrase 1", "trigger phrase 2".
---

Example: Forked/Isolated Skill (Use Sparingly)

Only use forked context for tasks that truly need isolation:

---
name: heavy-analysis
description: Run deep analysis in isolation
context: fork
agent: Explore
allowed-tools: Read, Grep, Glob
---

When to Use context: fork

Use forked context when:

  • The task is completely self-contained and needs no prior context
  • You want to run a long-running analysis without blocking the main conversation
  • The task should not affect or be affected by the current conversation state

Do NOT use forked context when:

  • The skill needs to reference previous conversation messages
  • The skill modifies files that the user is actively working on
  • The skill needs to ask follow-up questions or clarifications

Quick Start

Creating a New SOP

  1. Choose Location: /docs/processes/{category}/ where category is:

    • development/ - Development workflows
    • design/ - Design and research processes
    • tests/ - Testing procedures
    • code_review/ - Review processes
    • troubleshooting/ - Diagnostic guides
    • git/ - Version control processes
  2. Name the File: Use snake_case with descriptive name

    {category}_{process_name}_process.md
    

    Examples: code_review_process.md, testing_process.md

  3. Use the Template: Copy SOP_TEMPLATE.md and fill in sections

  4. Review Checklist:

    • All placeholder text replaced
    • Metadata section complete
    • At least 2 phases defined
    • Quality checks added
    • Troubleshooting section populated
    • Related documents linked

Process Workflow

Flow Diagram

[Identify Need] --> [Choose Category & Name] --> [Copy Template]
       |
[Fill Metadata] --> [Define Phases] --> [Add Examples]
       |
[Add Troubleshooting] --> [Link Related Docs] --> [Review]
       |
[Publish] --> [Consider Skill Conversion]

Phase Summary

PhaseObjectiveDeliverableDuration
1. PlanningIdentify scope and structureOutline with phases15-30 min
2. DraftingWrite core contentComplete draft1-2 hours
3. ReviewVerify completenessFinalized SOP30-60 min
Total2-4 hours

Phase 1: Planning

Objective: Define the scope, structure, and key phases of the SOP Duration: 15-30 minutes

Step 1.1: Identify the Process

  1. Answer these questions:

    • What problem does this process solve?
    • Who will use this process?
    • What is the expected outcome?
    • How often will it be used?
  2. Define the scope:

    • What is covered by this SOP?
    • What is explicitly NOT covered?
    • What prerequisites exist?

Step 1.2: Structure the Phases

  1. Break the process into 2-5 logical phases
  2. Each phase should:
    • Have a clear objective
    • Produce a deliverable
    • Have estimable duration
    • Be independently verifiable

Deliverable

Output: Process outline with phases defined

Quality Check:

  • Process has clear purpose
  • Scope is well-defined
  • 2-5 phases identified
  • Each phase has clear output

Phase 2: Drafting

Objective: Write the complete SOP content Duration: 1-2 hours

Step 2.1: Fill Metadata

# [Process Name] Process Guide

**Version**: 1.0.0
**Last Updated**: YYYY-MM-DD
**Status**: Active
**Document Type**: Process Guide
**Project**: {Project Name}

Step 2.2: Write Overview Section

Include:

  • Purpose (2-3 sentences)
  • Scope (Covered / NOT Covered)
  • When to Use (ALWAYS / SKIP)
  • Summary (2-3 paragraphs)

Step 2.3: Write Phase Sections

For each phase:

## Phase N: [Phase Name]

**Objective**: [Clear statement]
**Duration**: [Time estimate]

### Step N.1: [Step Name]
1. Action 1
2. Action 2

### Deliverable
**Output**: [What this phase produces]

**Quality Check**:
- [ ] Criterion 1
- [ ] Criterion 2

Step 2.4: Add Supporting Sections

  • Quick Reference (commands, file locations)
  • Troubleshooting table
  • Related Documentation links
  • Changelog

Deliverable

Output: Complete SOP draft

Quality Check:

  • All sections filled
  • No placeholder text remains
  • Commands are correct
  • Links are valid

Phase 3: Review

Objective: Verify completeness and quality Duration: 30-60 minutes

Step 3.1: Self-Review

Use the SOP Quality Checklist:

Structure:

  • Has YAML-like metadata header
  • Has Overview section with Purpose, Scope, When to Use
  • Has Process Workflow with diagram
  • Has 2+ Phase sections
  • Has Quick Reference section
  • Has Troubleshooting section
  • Has Related Documentation section
  • Ends with "End of Document" marker

Content:

  • Purpose is clear (answers "why")
  • Scope is defined (covered / not covered)
  • When to Use has ALWAYS and SKIP conditions
  • Each phase has objective, steps, deliverable, quality check
  • Commands are tested and correct
  • File paths are accurate

Style:

  • Uses consistent heading levels
  • Uses markdown formatting correctly
  • No spelling/grammar errors
  • Action items use imperative verbs

Step 3.2: Consider Skill Conversion

If this SOP should become a Claude Code skill:

  1. Create skill in .claude/skills/{skill-name}/
  2. Extract core workflow into SKILL.md
  3. Use progressive disclosure for details
  4. Add YAML frontmatter with triggers

Deliverable

Output: Finalized, reviewed SOP

Quality Check:

  • All checklist items pass
  • Ready for publication
  • Skill conversion assessed

SOP Quality Checklist

Required Sections

SectionRequiredPurpose
Title + MetadataYesVersion, date, status
OverviewYesPurpose, scope, when to use
Process WorkflowYesVisual flow, phase summary
Phase 1-NYesStep-by-step instructions
Quick ReferenceRecommendedCommands, file locations
TroubleshootingRecommendedCommon issues and solutions
Related DocumentationYesLinks to related SOPs
ChangelogRecommendedVersion history

Naming Conventions

TypeConventionExample
SOP File{category}_{name}_process.mdcode_review_process.md
Skill Folder{prefix}-{name}review-code
Skill FileSKILL.mdSKILL.md
Template File{NAME}_TEMPLATE.mdSOP_TEMPLATE.md

Status Values

StatusMeaning
DraftIn development, not ready for use
ActiveCurrent, approved for use
DeprecatedBeing phased out, use alternative
ArchivedHistorical reference only

Quick Reference

File Locations

TypeLocation
Process SOPs/docs/processes/{category}/
Setup Docs/docs/SETUP_*.md
Reference Docs/docs/*.md
Skills/.claude/skills/

Template Location

Full SOP template: SOP_TEMPLATE.md

Skill Names (New Convention)

NameDomainLocation
create-skillMeta/documentationshared/
dev-featureDevelopment workflowsshared/
design-researchDesign/researchshared/
git-prGit operationsshared/
setup-devEnvironment setup{domain}/
setup-envEnvironment variables{domain}/
testTesting{domain}/
review-codeCode review{domain}/
deployDeployment{domain}/
helpTroubleshooting{domain}/

Troubleshooting

IssueSymptomsSolution
Template not foundCan't locate SOP_TEMPLATE.mdCheck .claude/skills/shared/create-skill/SOP_TEMPLATE.md
Unclear scopeProcess covers too muchBreak into multiple SOPs
Too detailedSOP over 500 linesUse progressive disclosure with linked files
Missing examplesUsers struggle to followAdd concrete examples for each phase
Outdated contentProcess has changedUpdate SOP and increment version

When to Escalate

  • Process spans multiple teams
  • Requires external tool integration
  • Conflicts with existing SOPs

Related Skills

SkillPurposeWhen to Use
/dev-featureFeature development workflowCreating feature documentation
/design-researchResearch and design processResearching before documenting

Note: Skill paths (/skill-name) work after deployment. In the template repo, skills are in domain folders.


End of SOP

Skills Info
Original Name:create-skillAuthor:eleva