n8n Multi-Instance Targeting
Safely select, verify, and switch the n8n instance targeted by an MCP session.
The documentation explicitly requires instance discovery, post-switch confirmation, and re-verification before credential writes; it also provides fail-closed handling for INSTANCE_AMBIGUOUS and treats credentials as high-risk data. Points are deducted because ordinary reads and writes, and an explicit switch to the wrong instance, can still misroute silently; no general rollback or undo procedure is provided; sensitive-data handling is partly delegated to an unreviewed companion skill; and publisher identity is unverified.
SKILL.md gives detailed tool modes, error codes, recovery procedures, and a parallel-call ordering warning, making the happy path plausible. Points are deducted because there is no committed test suite or CI evidence covering key paths, and the README says there are five golden rules while the skill lists six, weakening instruction consistency and diagnostic confidence.
Activation triggers, the multi-instance scenario, the single-instance non-fit boundary, and representative use cases are clearly described, including misroutes, copying, and credential operations. Points are deducted because Chinese-language behavior and mainland-China network reachability are not addressed, and compatibility across n8n-mcp versions or tool variants is not sufficiently bounded.
The documentation is well organized and includes core rules, tool shapes, a recovery table, quick reference, integrations, version 1.0.0, compatibility notes, and an MIT license. Points are deducted because there is no changelog, explicit maintenance owner, or update path; the README contains a rule-count inconsistency; and repository-level testing claims do not establish coverage for this selected skill.
The skill directly supplies the sequence for listing, switching, confirming, retrying, and copying across instances, plausibly addressing its core targeting problem and reducing mistaken recreation or credential misplacement. Points are deducted because no execution was performed and no reproducible representative outputs or third-party results are supplied; workflow and credential copying still depend on details owned by other skills.
The files provide concrete call shapes, error codes, response fields, recovery tables, and version information, giving limited auditability. Points are deducted because claims about real responses and production testing are not independently corroborated by a committed test suite, CI workflow, or third-party evidence within the supplied material, and static review cannot verify alignment with the actual n8n-mcp implementation.
- An explicit switch to the wrong instance is not blocked by INSTANCE_AMBIGUOUS; verify current before credential writes and preferably before ordinary writes as well.
- The selected skill contains no automated test or CI evidence; validate switching, deleted-instance fallback, and misroute recovery in an isolated staging instance before deployment.
- Chinese-language support and mainland-China network reachability are unspecified; confirm that the n8n-mcp service and MCP connection are accessible in the intended environment.
What does this skill do, and when should you use it?
This is one standalone Agent Skill in the n8n-skills repository, focused on n8n-mcp accounts that can reach multiple n8n instances. It instructs an assistant to discover available instances, switch the session by name, and verify the target before credential writes or destructive edits. It also covers NOT_FOUND responses, unfamiliar data, empty reads, and INSTANCE_AMBIGUOUS credential failures caused by misrouting. The Skill handles instance targeting and recovery; it does not replace other Skills that document workflow or credential field operations.
Guides the assistant to call n8n_instances in list mode and read current, default, and available instances; call switch mode to bind the MCP session to a named instance; keep switching separate from dependent operations; re-check current before credential creation, updates, deletions, or destructive workflow edits; recover from UNKNOWN_INSTANCE, INSTANCE_AMBIGUOUS, NOT_FOUND, empty data, and unexpected health-check instance names; and copy workflows or credentials between explicitly selected instances.
- An operator manages production and staging n8n instances and needs to confirm the active environment before making changes.
- A consultant serves several teams or clients with separate n8n instances and must switch targets by instance name.
- A user receives NOT_FOUND for a workflow or datatable that is known to exist and needs to diagnose a wrong-instance route.
- An operator is about to create, update, or delete credentials and must verify the destination before writing secrets.
- A credential request returns INSTANCE_AMBIGUOUS and the user needs the session-specific recovery sequence.
What are this skill's strengths and limitations?
- Defines a concrete discover-switch-verify sequence that reduces silent routing mistakes.
- Treats credential writes as high risk and explains the INSTANCE_AMBIGUOUS fail-closed behavior.
- Provides recovery procedures for NOT_FOUND, empty reads, unexpected health-check targets, and fallback to the default instance.
- Documents the ordering needed to copy workflows or credentials between instances.
- Requires an n8n-mcp environment with the n8n_instances tool; single-instance accounts should ignore this Skill.
- Focuses on instance selection and recovery rather than complete CRUD schemas for other n8n tools.
- The source does not document a standalone execution path outside a compatible Skill client and MCP session.
- An explicit switch to the wrong instance can still write silently, so the operator must verify the target.
How do you install this skill?
Install it as part of the n8n-skills collection. For Claude Code, the documented recommended command is: /plugin install czlonkowski/n8n-skills. For manual installation, run git clone https://github.com/czlonkowski/n8n-skills.git and copy n8n-skills/skills/* to ~/.claude/skills/. The collection requires an installed and configured n8n-mcp MCP server and a configured .mcp.json. For Claude.ai, download the skills/n8n-multi-instance folder and upload it through Settings → Capabilities → Skills.
How do you use this skill?
Use it when n8n_instances is available, when the user mentions multiple n8n environments, or when calls return NOT_FOUND, unfamiliar or empty data, or INSTANCE_AMBIGUOUS. Example prompts include: "Switch to the prod n8n instance, then inspect its workflows" and "A credential write returned INSTANCE_AMBIGUOUS; recover safely." Execution depends on the n8n tools available in the current MCP session.