N8n:create Skill

Creating skills

Skills are markdown (plus optional scripts) that teach the agent a focused workflow. Keep SKILL.md short—the context window is shared with chat, code, and other skills.

Where skills live

| Location | When to use |

|----------|-------------|

| .agents/skills/<name>/ | Default for n8n: team-shared, versioned, agent-neutral source. |

| .claude/plugins/n8n/skills/<name>/ | Claude-specific override, or a generated symlink to .agents/skills/<name>/. |

| .opencode/skills/<name>/ | OpenCode-specific override only. Shared skills stay in .agents/skills/<name>/. |

| ~/.claude/skills/<name>/ | Personal skill for Claude Code across all projects. |

| ~/.config/opencode/skills/<name>/ | Personal skill for OpenCode across all projects. |

| ~/.cursor/skills/<name>/ | Optional personal skill for Cursor only, global to your machine. |

Do not put custom skills in ~/.cursor/skills-cursor/—that is reserved for Cursor’s built-in skills.

Prefer .agents/skills/ for anything that should match how the rest of the team works. Run pnpm sync:skill-links after adding or removing shared skills.

Before you write: gather requirements

Ask (or infer) briefly:

• Purpose — one concrete task or workflow.
• Triggers — when should the agent apply this skill?
• Gaps — what does the agent *not* already know (project rules, URLs, formats)?
• Outputs — templates, checklists, or strict formats?
• Examples — follow an existing shared skill in .agents/skills/ if one fits.

Ask the user in plain language when you need more detail.

File layout

skill-name/
├── SKILL.md       # required
├── reference.md   # optional — detail the agent reads only if needed
├── examples.md    # optional
└── scripts/       # optional

Frontmatter (required)

---
name: n8n:skill-name      # n8n:<name> — lowercase, hyphens, max 64 chars
description: >-         # max 1024 chars, non-empty — see below
  ...
---

Name — shared n8n skills use the n8n:<name> form so Claude Code namespaces

them under the n8n plugin (invoked as /n8n:<name>). The <name> part must

match the skill's directory name.

Description (discovery is everything — third person, WHAT + WHEN, trigger words):

• Good: Extracts tables from PDFs and fills forms. Use when the user works with PDFs, forms, or document extraction.
• Bad: Helps with documents or I can help you with PDFs

Authoring rules

• Concise — assume the model is capable; only add non-obvious domain or project facts.
• Progressive disclosure — essentials in SKILL.md; long reference in reference.md. Link one level deep from SKILL.md.
• Prefer one default — e.g. one library or one workflow; add an escape hatch only if needed.
• Stable wording — one term per concept; avoid dated “until month X” notes unless you tuck legacy bits behind a short “Deprecated” note.
• Paths — forward slashes only (scripts/foo.py).

Rough size: aim for well under ~200 lines in SKILL.md; if it grows, split detail out.

Scope: one job per skill (and parent skills)

• Single responsibility — one primary workflow or decision tree per skill. If triggers and steps diverge a lot (e.g. “create issue” vs “create PR” vs “full ticket → PR flow”), split into smaller dedicated skills.
• Prefer small + compose — two or three focused skills keep irrelevant detail out of context until needed. A parent (orchestrator) skill can say *when* to follow each child workflow and link to their SKILL.md; avoid pasting full child content into the parent.
• When one large skill is OK — a single end-to-end flow that always runs together and shares one tight checklist;

MCPs, CLI tools, and other skills

• Prefer CLI and repo commands when they solve the same problem — agents handle them well and they usually add less scaffolding noise to context than MCP tool discovery and schemas. Examples: gh for PRs/issues, pnpm scripts from AGENTS.md.
• MCPs are optional per user — not everyone has the same servers enabled. If a skill requires a specific MCP to work as written, say so explicitly:
• Put a hint in the frontmatter description (e.g. “Requires Linear MCP for …”) so mismatches are obvious early.
• Add a short Prerequisites (or Requirements) block near the top: which integration, what it is used for, and a fallback (e.g. web UI, gh, or “ask the user to paste …”) when it is missing.
• Referencing other skills — use the harness-visible invocation name (e.g. n8n:create-issue where namespacing is available, otherwise create-issue). For human-readable links, give the canonical path from the repo root (e.g. .agents/skills/create-issue/SKILL.md). From a sibling folder, a relative link works too: create-issue. Parent skills should delegate steps instead of duplicating long procedures.

Patterns (pick what fits)

• Template — give the exact output shape (markdown/code blocks).
• Checklist — numbered or - [ ] steps for multi-step work.
• Branching — “If A → …; if B → …” at the top of a workflow.
• Scripts — document run commands; say whether to execute or read the script.

Workflow: create → verify

• Name + description — hyphenated name; description with triggers.
• Outline — minimal sections; link optional files.
• Implement — SKILL.md first; add reference.md / scripts/ only if they save tokens or reduce errors.
• Check — third-person description; terminology consistent; no duplicate encyclopedic content the model already knows.

Anti-patterns

• Verbose tutorials (“what is a PDF”) inside the skill.
• Many equivalent options with no default.
• Vague names (helper, utils).
• Deep chains of linked files.
• Assuming an MCP or tool is present without stating it or offering a fallback.
• One oversized skill that mixes unrelated workflows instead of smaller skills + a thin parent.

Quick example stub

---
name: n8n:my-workflow
description: Does X using project convention Y. Use when the user asks for X or mentions Z.
---

# My workflow

1. …
2. …

## Output format

Use a fenced code block for the exact shape reviewers should see.

## More detail
See [reference.md](reference.md) if edge cases matter.