add fabric security skill to be included in the plugin

.github/plugin/plugins/rayfin/plugin.json

@@ -15,5 +15,8 @@
     "fabric",
     "typescript"
   ],
-  "skills": ["skills/rayfin-getting-started"]
+  "skills": [
+    "skills/rayfin-getting-started",
+    "skills/security-best-practices"
+  ]
 }

.github/plugin/plugins/rayfin/skills/rayfin-getting-started/SKILL.md

@@ -74,3 +74,18 @@ loading the in-project skill: `create-rayfin` creates a child project directory
 `--project-name`, slugified), so `cd` into it; an in-place `rayfin init` scaffolds in the
 current directory, so you're already there. Once at the project root, load its
 `.agents/skills/rayfin/SKILL.md`.
+
+## Security best practices
+
+Build security into the app from the start:
+
+- Request only the Fabric data and connector permissions the app needs.
+- Filter sensitive data server-side and return only the rows and columns required by the UI.
+- Keep secrets, tokens, and environment-specific identifiers out of client code, URLs, and logs.
+- Avoid rendering PII or detailed errors in the browser unless the user is authorized to see them.
+- Remove debug output and verify production configuration before deployment.
+
+Before deployment, run the
+[Fabric App Security Review](../security-best-practices/SKILL.md) skill to
+check for sensitive data exposure, semantic model risks, unsafe configuration, excessive
+permissions, and information leakage.

.github/plugin/plugins/rayfin/skills/security-best-practices/SKILL.md

@@ -0,0 +1,33 @@
+---
+name: fabric-app-security-review
+description: Review a Fabric App for sensitive data exposure, semantic model security issues, connector risks, and deployment readiness.
+---
+
+# Fabric App Security Review
+
+When invoked, perform a security review of the target Fabric App codebase. Focus on whether the app exposes sensitive data to the browser, logs private information, over-fetches from Fabric items, or hardcodes environment-specific configuration.
+
+## Review Steps
+
+1. Review semantic model usage for full-table reads, raw transaction access, unfiltered queries, excessive columns, or queries that return more data than the UI requires.
+2. Review React or frontend UI code for PII rendered in the DOM, sensitive props passed between components, or restricted data hidden only with client-side logic.
+3. Review API calls for tokens, secrets, workspace IDs, item IDs, tenant IDs, or sensitive query parameters in URLs, request bodies, or logs.
+4. Review configuration files such as .env files, manifest.json, app settings, deployment files, and environment configs for hardcoded secrets or identifiers.
+5. Review logging and error handling for console.log, console.debug, verbose error objects, stack traces, and debug-only pages that could leak sensitive information.
+6. Review connector and Fabric item usage for broad permissions, unnecessary metadata exposure, or environment-specific values that should be resolved securely at runtime.
+
+## Focus Areas
+
+- PII exposure: names, emails, phone numbers, employee data, customer identifiers, salary, financial, or account data.
+- Excessive semantic model access: full table retrieval, SELECT-style patterns, raw rows, or large dimensions when measures or aggregates are safer.
+- Workspace, item, and tenant ID exposure: identifiers visible in client bundles, URLs, logs, or error messages.
+- Client-side filtering: fetching all records and filtering in the browser instead of enforcing least privilege before data reaches the client.
+- Debug artifacts: console output, stack traces, temporary debug UI, or commented code that exposes implementation details.
+
+## Output Format
+
+Return findings grouped by Critical, High, Medium, and Low severity. For each finding, include location, issue, why it matters, evidence, recommended fix, and suggested Copilot prompt to remediate it.
+
+## Remediation Guidance
+
+Do not only describe the problem. Recommend a concrete fix, such as moving the filtering server-side, replacing raw data with curated measures, removing logs, moving configuration out of client code, narrowing connector scopes, or using safe error messages.