Frontegg Coding-Agent Skills

29 trigger-activated skills for Frontegg integration — one canonical source, every major IDE

Author once. Compile to every IDE. Trigger by intent, not memorized commands.

Works with Cursor · VS Code · Windsurf · JetBrains · Zed · Cline · Roo Code · Aider · Claude Code · Codex

Table of contents

• Highlights
• Skills
• Quick start
• Supported IDEs
• Output tiers
• CLI
• Architecture
• Authoring
• Linter
• FAQ & troubleshooting

Highlights

Single canonical source

29 skills authored once in canonical/. Compiled into 3 tiers (Universal / Standard / Rich) for every IDE format.

10+ IDEs supported

Cursor, VS Code, Windsurf, JetBrains, Zed, Cline, Roo, Aider, Claude Code, Codex — same content, right format per tool.

Trigger by intent

Each skill activates on natural language ("how do I add passkeys?", "fix my deep link") — no memorized commands.

Two personas in one body

Skills cover both the developer integrating Frontegg's SDK and the operator managing the tenant. Overlap skills split into Read Side / Write Side sections.

Fabrication-resistant

Linter denylist hard-fails on parameter names LLMs have invented in the past (passkeys_enabled, mfa_options, eventType, etc.).

Source-cited

Each skill cites the canonical Frontegg source — official docs, SDK repos, and API references — so the agent always has authoritative context.

Skills

| Skill | Target | Triggers | |---|---|---| | frontegg-react | React | "integrate Frontegg in React", "FronteggProvider setup" | | frontegg-nextjs | Next.js | "Frontegg in Next", "middleware auth Next.js" | | frontegg-angular | Angular | "Frontegg Angular module" | | frontegg-vue | Vue | "Frontegg in Vue", "Vue auth setup" | | frontegg-react-native | React Native | "Frontegg RN", "React Native hosted login" | | frontegg-ios | iOS (Swift) | "integrate Frontegg in iOS", "FronteggWrapper", "Sign In button does nothing" | | frontegg-android | Android (Kotlin/Java) | "Frontegg on Android", "FronteggApp.init", "hosted login never appears" |

| Skill | Domain | |---|---| | frontegg-getting-started | Front-door routing; narrows intent into a specific skill | | frontegg-auth-flows | Hosted login, embedded login, passwordless | | frontegg-passkeys | WebAuthn, Associated Domains, asset links | | frontegg-multi-tenant | Tenant model, switching, isolation | | frontegg-multi-app | Application registration, agent apps, user assignment | | frontegg-user-pools | User-pool architecture | | frontegg-sso-guide | SAML, OIDC, social providers | | frontegg-scim | SCIM provisioning | | frontegg-rbac | Roles, permissions, permission categories | | frontegg-entitlements | Feature gating, plans, feature flags, ReBAC | | frontegg-prehooks | Pre-event interceptors, JWT claim injection | | frontegg-webhooks | Outbound event notification | | frontegg-security | Token-based access, API tokens, best practices | | frontegg-tenant-config-playbook | Tenant-wide policies via Management API (MFA, sessions, SSO, branding, lockout) | | frontegg-api-tokens | Tenant tokens, OAuth client-credentials, personal access tokens | | frontegg-deep-link-troubleshooting | OAuth callback / Associated Domains / asset links debug | | frontegg-styling-and-theming | Login box theming, Admin Portal branding, CSS fixes | | frontegg-integrations-management | Vendor vs. Frontegg integrations, agent assignment |

| Skill | Purpose | |---|---| | frontegg-monitoring | Audit logs, user activity, session state, MFA forensics | | frontegg-backend | Backend integration patterns | | frontegg-agent-safety | Cross-cutting safeguards for any agent operating on a Frontegg project | | frontegg-troubleshoot | Cross-skill generic debug |

All 29 pass python3 scripts/lint-skills.py.

Quick start

This repo installs standalone — skills are pure markdown playbooks that work with any AI coding assistant.

⚠️ Pre-release. @frontegg/coding-agent-skills is not yet published to npm. Until then, use the manual install below (clone + build + copy). The npx command will work once the package ships.

# Coming soon — once published to npm:
npx @frontegg/coding-agent-skills init

The CLI detects your framework (@frontegg/react, @frontegg/nextjs, etc.) and IDE, then installs the right skill files into your project.

Manual install (works today)

git clone https://github.com/frontegg/coding-agent-skills.git
cd coding-agent-skills
npm install && npm run build      # compiles canonical/ → build/skills/

# Cursor
cp -r build/skills/* ~/.cursor/skills/

# Claude Desktop (macOS)
cp -r build/skills/* ~/Library/Application\ Support/Claude/skills/

# Claude Code
for s in build/skills/*/; do claude skills install "$s"; done

Restart your client. Trigger phrases from each skill's description field will now activate the skill.

Onboarding your first app

Once installed, open your AI coding assistant and walk through these steps in order:

1. "I'm new to Frontegg — help me get started" — the frontegg-getting-started skill orients you and routes to the right next step
2. "Set up Frontegg login in my app" — auto-detects your framework and walks you through provider setup, hooks, and the first working login
3. "Add a protected page that requires login" — shows you how to gate routes and check auth state using Frontegg hooks
4. "Customize my login page look and feel" — the frontegg-styling-and-theming skill handles branding, colors, layout, and CSS

Once you're up and running, explore deeper features:

• "Add SSO configuration" — enterprise SAML/OIDC setup
• "How do I enable multi-tenancy?" — tenant model, switching, and isolation
• "Add entitlements to gate a feature" — plans, feature flags, and ReBAC

No portal credentials or extra tooling required — skills teach your AI assistant Frontegg patterns, syntax, and best practices purely from markdown.

Supported IDEs

| IDE | Output format | Tier | |---|---|:---:| | Cursor | .cursor/rules/*.mdc | Rich | | VS Code / Copilot | .github/instructions/ | Standard | | Windsurf | .windsurfrules | Standard | | JetBrains | .github/instructions/ | Standard | | Zed | AGENTS.md | Universal | | Cline | .clinerules/ | Standard | | Roo Code | .roo/ | Standard | | Aider | AGENTS.md | Universal | | Claude Code | .agents/skills/*/SKILL.md | Rich | | OpenAI Codex | AGENTS.md | Universal |

Output tiers

Each skill compiles to three detail levels. The CLI picks the best fit per IDE; you can override with --tier.

Universal

AGENTS.md — single file, ~150 lines. Condensed. Works with any IDE that reads agents-flavored prompts.

Standard

Per-tool format files (.github/instructions/, .windsurfrules, .clinerules/, CLAUDE.md). Medium detail.

Rich

.cursor/rules/*.mdc + .agents/skills/*/SKILL.md — full canonical content with glob activation + progressive loading.

CLI

npx @frontegg/coding-agent-skills init          # auto-detect IDE + SDK, install
npx @frontegg/coding-agent-skills init --all    # install all SDK skills
npx @frontegg/coding-agent-skills update        # update to latest skill versions
npx @frontegg/coding-agent-skills list          # show installed skills
npx @frontegg/coding-agent-skills add <skill>   # add a specific skill
npx @frontegg/coding-agent-skills remove        # remove all installed skill files
npx @frontegg/coding-agent-skills doctor        # validate installation integrity

Flags

--ide <cursor|vscode|windsurf|jetbrains|cline|roo|claude-code|codex|all>
--sdk <react|nextjs|angular|vue|react-native|ios|android|all>
--tier <universal|standard|rich>
--dry-run

Architecture

coding-agent-skills/
├── canonical/          # source of truth (authored Markdown, one per skill)
│   ├── sdk/            # per-framework integration skills (7)
│   ├── platform/       # Frontegg features + admin workflows (17)
│   └── operations/     # cross-cutting observability + troubleshooting (3)
│
├── generators/         # canonical → per-IDE compilers
├── templates/          # per-IDE wrapper templates
├── scripts/
│   └── lint-skills.py  # schema + tool-reference + denylist validation
├── bin/
│   └── cli.js          # npx @frontegg/coding-agent-skills entry
└── .github/workflows/  # CI: lint-skills + build verification

Authoring

Every canonical skill must:

1. Use the YAML frontmatter schema (name, description, tags, category, targets, priority, visibility, license, metadata)
2. Start its description with "Use when..." — never summarize the skill's workflow
3. Contain a ## When to Use This Skill block (Must Use / Recommended / Skip When / Decision)
4. Cite at least one canonical Frontegg URL (docs.frontegg.com, GitHub frontegg/* SDK repos)
5. End with a ## Reference section
6. Never use a parameter name from the FABRICATED_FIELDS denylist (see scripts/lint-skills.py)

CI runs npm run build + python3 scripts/lint-skills.py on every PR. PRs introducing a new fabricated field name auto-fail.

Adding a skill

# 1. Author in canonical/<category>/<skill-name>.md
# 2. Lint
python3 scripts/lint-skills.py

# 3. Build
npm run build

# 4. Verify your skill appears in build/skills/
ls build/skills/frontegg-<your-skill>

Linter

python3 scripts/lint-skills.py

Validates every canonical/*/*.md against:

1. YAML frontmatter present with required fields
2. name matches filename
3. category matches parent directory and is one of sdk / platform / operations
4. description contains "Use when"
5. Body ≥ 800 chars
6. Body cites at least one canonical Frontegg URL
7. Required section: ## Reference
8. Encouraged (warning, not error): ## When to Use This Skill
9. No fabricated parameter names — denylist hard-fails on tokens we've caught LLMs inventing (passkeys_enabled, mfa_options, lockout_duration_minutes, eventType, etc.)

FAQ & troubleshooting

Built for engineering teams integrating Frontegg. Every skill is a real playbook. Every parameter name traces to the actual tool schema. Every diff cites a canonical Frontegg source.