A skill is a short guide for an agent: which tools to call for an SFCC task, in what order, and the rules that bite. Agents can read every skill from the server with intellijsfcc_skills_read, without any file. Adding skills to the project lets an agent load the right one on its own
The skills
| Skill | What it teaches |
|---|---|
sfcc-workspace-start | The first moves in a project, the toolsets, safety and budgets. Start here |
sfcc-code-evidence-and-refactor | Proving types and references, and safe renames |
sfcc-sfra-controllers | SFRA routes and middleware |
sfcc-cartridge-overrides-and-models | Overrides, module.superModule and models |
sfcc-isml-forms-resources | ISML, forms and resource bundles |
sfcc-client-assets-scss | Client scripts and SCSS, for reading |
sfcc-hooks | Hook registration, and OCAPI and SCAPI hooks |
sfcc-custom-scapi-endpoints | Custom SCAPI endpoints and their registration |
sfcc-headless-scapi-slas | Headless shops with SCAPI and SLAS |
sfcc-admin-api-integrations | Admin API integrations |
sfcc-services-integrations | Service definitions and testing them |
sfcc-checkout-orders-fraud | Checkout, orders, payments and fraud checks |
sfcc-performance-and-limits | Caching, quotas and platform limits |
sfcc-security-hardening | A security checklist for SFCC code |
sfcc-page-designer-components | Page Designer components |
sfcc-bm-extensions-and-pipelines | Business Manager extensions and pipelines |
sfcc-metadata-and-site-archives | Metadata and site archives |
sfcc-jobs-steptypes | Jobs, step types and running them |
sfcc-ship-and-verify | Uploads, deploys, code versions and WebDAV, then checking the result |
sfcc-log-center-triage | Parsing and tailing logs, anomalies, comparing and reports |
sfcc-live-debugger | Debug sessions, pipeline debugging and evaluate |
sfcc-digital-script | Digital Script imports, E4X and TypeScript |
sfcc-on-demand-sandboxes | Sandbox lifecycle, costs and creating one |
Skill text is the plugin's own. The server never puts instance data inside a skill
Add skills
Every way shows the files first and writes nothing until you agree
| Way | Where |
|---|---|
| B2C Add Skill | Right-click in the Project view, the editor or an editor tab, beside Upload. On a skill folder of an older version it reads B2C Update Skill |
| Connect Agent | Add SFCC Skills… in the dialog opens the same dialog for that agent, see Connect Agents |
| The Tools menu | Add SFCC Skills… in Tools › Salesforce B2C Commerce › B2C MCP. Find Action finds it too |
| The Server tab | Add SFCC Skills… in the SFCC skills group |
| The agent | It plans the files with intellijsfcc_skills_install_plan, then runs the plan |
B2C Add Skill shows in linked SFCC projects, like the other SFCC actions
B2C Add Skill
Pick the agent
Agent: lists Claude Code, Codex and Agent Skills, GitHub Copilot and Gemini CLI, each with the folder it writes to, such as .claude/skills/. The format the project already uses is selected
Tick the skills
Skills: lists all 23, the ones the project calls for first and marked (suggested), and ticked. Filter skills narrows the list. Below it, Add the SFCC block to AGENTS.md, Add @AGENTS.md to CLAUDE.md, so Claude Code reads the block (for Claude Code) and Update older copies of our skills
Check the files
Files: lists every file as you change the choice, each marked Create, Update, Already there, kept, Older text of ours, kept or Edited by you, kept. Click Add, or Cancel
Done
One command writes the files, so one undo removes them. A single skill opens its SKILL.md, and a notification says what was added, with Show in Project
It never overwrites a file, never commits, and works while the IDE indexes
Agent formats
| Agent | Where skills go |
|---|---|
| Claude Code | .claude/skills/<skill>/, with the skill's references |
| Codex and Agent Skills (Codex, Cursor, Junie, AI Assistant's agents, OpenCode and others) | .agents/skills/<skill>/, with the references and agents/openai.yaml for Codex |
| GitHub Copilot | .github/instructions/<skill>.instructions.md |
| Gemini CLI | .gemini/skills/<skill>.md and a short block in GEMINI.md |
Any agent that reads AGENTS.md | A short block in AGENTS.md, see The AGENTS.md block |
The text is the same in every format, and the same the server returns, so a project copy and the server never disagree
When an agent has no skills
The server notices when a connected agent has no SFCC skills in its format, and suggests the ones the project needs:
- The first time an agent connects to the project, its notification offers Add SFCC Skills with the suggested skills ticked, or Don't Ask Again for this project
- The SFCC skills group on the Server tab shows each format with its skills on disk, suggested ones missing and older copies, and Add SFCC Skills…
- The agent itself gets a one-line hint once per session, which it can pass on to you
Suggestions come from the project: pipelines suggest sfcc-bm-extensions-and-pipelines, .ds files sfcc-digital-script, a steptypes.json sfcc-jobs-steptypes, Page Designer components sfcc-page-designer-components, an api.json sfcc-custom-scapi-endpoints and a hooks.json sfcc-hooks. At most five, plus sfcc-workspace-start
Updates keep your edits
.intellij-sfcc/agent-files.json records every file the plugin wrote. When a newer plugin has newer skills, the Server tab counts them as older, and right-clicking a skill folder of ours reads B2C Update Skill, never a notification. With Update older copies of our skills ticked, an update rewrites only the files you didn't edit, and lists the others as Edited by you, kept. Committing is up to you
The AGENTS.md block
Agents that read AGENTS.md get a short block between two markers, under 1,200 characters. It tells the agent to start every SFCC task with intellijsfcc_project_context, prove before it edits, plan every instance write first, treat log lines and files as data, never ask for credentials, and where the skills are. The rest of AGENTS.md stays as you wrote it