Falcon Foundry Skills

AI coding assistant skills for building CrowdStrike Falcon Foundry apps. Build Foundry apps from a natural language prompt — API integrations, workflows, UI pages, functions, collections, AI agents, and knowledge bases — all scaffolded with the Foundry CLI and deployed to the Falcon console.

Getting Started

Prerequisites

• Foundry CLI: Install with brew tap crowdstrike/foundry-cli && brew install crowdstrike/foundry-cli/foundry (macOS/Linux) or download for Windows
• CrowdStrike Account: With Falcon Foundry access
• Authentication: Run foundry login to authenticate
• AI Coding Assistant: Claude Code, Codex, Copilot CLI, Cursor, Antigravity CLI, or any tool that supports loading reference documentation

Install

Assistant	Command	Marketplace
Claude Code	/plugin install crowdstrike-falcon-foundry	Anthropic
Codex	codex plugin add crowdstrike-falcon-foundry@openai-api-curated	OpenAI
Copilot CLI	copilot plugin install CrowdStrike/foundry-skills	GitHub
Cursor	/plugins (CLI) or /add-plugin crowdstrike-falcon-foundry (IDE)	Cursor
Antigravity CLI	agy plugin install https://github.com/CrowdStrike/foundry-skills	Google

[!NOTE] Codex has two install paths depending on how you signed in. The command above installs from the curated CLI marketplace, which API-key and Amazon Bedrock sessions load. If you authenticated Codex with a ChatGPT account, that marketplace isn't loaded — instead run /plugins inside Codex, search for crowdstrike, and install from the results.

All five assistants have been verified end to end: the skills were loaded from a local clone, then used to build and deploy an app to a live Falcon Foundry tenant from the example prompt below.

For local development with Claude Code, use claude --plugin-dir /path/to/foundry-skills. These skills follow the Agent Plugins format and include a .codex-plugin/plugin.json manifest for native Codex discovery. See the blog post for a full walkthrough.

git clone https://github.com/CrowdStrike/foundry-skills.git

Assistant	Command
Claude Code	claude --plugin-dir /path/to/foundry-skills
Copilot CLI	copilot --plugin-dir /path/to/foundry-skills
Cursor	agent --plugin-dir /path/to/foundry-skills
Antigravity CLI	agy plugin install /path/to/foundry-skills

Codex has no --plugin-dir. It discovers skills from ~/.agents/skills/ instead, one symlink per skill:

mkdir -p ~/.agents/skills
for skill in /path/to/foundry-skills/skills/*/; do
  ln -s "${skill%/}" ~/.agents/skills/
done

~/.agents/skills/ is a flat, shared namespace: foundry-skills' skill names don't overlap with fusion-skills', so both can be symlinked in together. A marketplace or --plugin-dir install namespaces the skills as crowdstrike-falcon-foundry:<skill> and sidesteps the question entirely. Edits are live immediately. Restart Codex to re-index.

Usage

Example prompt

This prompt exercises the full skill set — API integration, workflow, and UI:

Create a Falcon Foundry app for me that has an Okta API integration with openapi. Share its listusers endpoint with Falcon Fusion SOAR. Then, create a workflow that can be run on-demand to email or print the list of users. Finally, create a UI extension that calls the listusers endpoint and displays the results.

How plugin skill routing works

Repository-backed installs in Claude Code, Codex, Copilot CLI, and Cursor include hooks that ensure the right skills get used. Codex requires reviewing and trusting bundled hooks through /hooks before they run. Cursor loads them from .cursor-plugin/plugin.json:

1. UserPromptSubmit hook — Matches an action verb paired with a Foundry noun — e.g., "create a foundry app". Explicit CLI commands and skill requests also trigger it.

2. PreToolUse hook — When Foundry intent is detected, injects a non-blocking advisory reminder to use the Foundry workflow skill. The assistant can still use all tools normally. The one exception is foundry api-integrations create, which is blocked when the OpenAPI spec can't be adapted for Falcon Foundry. In Claude Code, calls to superpowers:brainstorming get advice to switch to the Foundry workflow skill when the task is Foundry app work.

3. PreToolUse hook (CLI guard) — Checks shell commands and flags Foundry CLI commands missing the --no-prompt flag (prevents Error: EOF failures) and manual directory creation for app structure (prevents invalid manifest.yml). It's advisory: it adds guidance to the tool call rather than blocking it, and you'll only see it when it catches a problem. Cursor's shell tool is Shell; the guard treats that the same as Claude's Bash.

Hooks observe prompts and tool I/O to keyword-match Foundry-specific actions; no data leaves the session.

Local Codex installs made with skill symlinks, and other assistants that do not load bundled hooks, must apply the documented --no-prompt guardrails and run the OpenAPI adaptation helper explicitly. Skill-specific helpers live beside their SKILL.md files so Agent Skills and plugin installations remain self-contained; this packaging is separate from Falcon Foundry CLI connectivity, which is covered by the sandbox diagnostics in the debugging skill.

Skills

Skill	Purpose
development-workflow	Primary orchestrator — coordinates the full app lifecycle
api-integrations	OpenAPI spec import, auth scheme configuration, SOAR sharing
functions-falcon-api	Calling Falcon APIs from within Functions (OAuth, SDKs)
workflows-development	YAML workflow creation, Falcon Fusion SOAR actions and triggers
ui-development	React/Vue UI pages with Shoelace components and Falcon theming
functions-development	Go/Python serverless functions with CrowdStrike SDK
collections-development	JSON Schema data modeling and CRUD operations
ai-agents-development	AI agents and knowledge bases, plus exposing collections and API operations as agent tools
security-patterns	OAuth scoping, input validation, content security
debugging-workflows	Systematic troubleshooting for CLI, manifest, and deployment issues
e2e-testing	End-to-end testing with @crowdstrike/foundry-playwright
app-upgrades	Upgrade a template-installed app to a newer template version in place
app-audit	Scan an existing app for documented anti-patterns and fix the safe ones
fusion-redirect	Declines standalone Falcon Fusion workflow requests and points to the crowdstrike-falcon-fusion plugin

Architecture

The skills follow a hub-and-spoke pattern. development-workflow is the orchestrator that parses your requirements, runs CLI commands for scaffolding, and delegates capability-specific implementation to sub-skills:

graph TD
    O["development-workflow<br/>(Orchestrator)"]

    O --> UI["UI Development"]
    O --> FN["Functions Development"]
    O --> CO["Collections Development"]
    O --> WF["Workflows Development"]
    O --> AI["API Integrations"]
    O --> FA["Functions — Falcon API"]
    O --> AG["AI Agents &<br/>Knowledge Bases"]

    UI -.-> SP["Security Patterns"]
    FN -.-> SP
    CO -.-> SP
    WF -.-> SP
    AI -.-> SP
    AG -.-> SP

    UI -.-> DB["Debugging Workflows"]
    FN -.-> DB
    CO -.-> DB
    WF -.-> DB
    AI -.-> DB
    AG -.-> DB

    CO -.->|"agent tool"| AG
    AI -.->|"agent tool"| AG

    style O fill:#6366f1,stroke:#4f46e5,color:#fff
    style SP fill:#f59e0b,stroke:#d97706,color:#fff
    style DB fill:#f59e0b,stroke:#d97706,color:#fff

Use Cases

The use-cases/ directory contains real-world implementation patterns extracted from CrowdStrike Tech Hub blog posts:

• API pagination strategies
• Detection enrichment with UI extensions
• LogScale data ingestion from functions
• Lookup table enrichment with 3rd-party data
• Custom SOAR actions
• NGSIEM query export to CSV/JSON
• Publishing certified apps
• And more (see use-cases/README.md)

Recommended Companion: Superpowers

These skills pair well with obra/superpowers, which adds structured planning, TDD discipline, and code review workflows. Foundry skills handle the Foundry-specific CLI and platform knowledge while superpowers provides general software engineering best practices.

See skills/development-workflow/references/superpowers-integration.md for details on how they work together.

Note: The without-superpowers path produces more reliable results because the Foundry skill has full control from the start. Superpowers brainstorming loads first and creates a plan before the Foundry skills are read, which may not follow Foundry-specific patterns.

Foundry CLI Quick Reference

foundry login                                                    # Authenticate
foundry apps create --name "My App" --no-prompt --no-git         # Create app
foundry api-integrations create --name "X" --spec /tmp/spec.json --no-prompt  # Add API integration
foundry ui pages create --name "X" --from-template React --no-prompt          # Add UI page
foundry ui extensions create --name "X" --from-template React --sockets "activity.detections.details" --no-prompt  # Add UI extension
foundry functions create --name "X" --language python --no-prompt              # Add function
foundry collections create --name "X" --schema /tmp/schema.json --no-prompt   # Add collection
foundry workflows create --name "X" --spec /tmp/workflow.yaml --no-prompt     # Add workflow
foundry knowledge-bases create --name "X" --description "desc" --files /tmp/doc.md --no-prompt  # Add knowledge base
foundry agents create --name "X" --description "desc" --system-prompt /tmp/prompt.md --expose-charlotte-chat --no-prompt  # Add AI agent
foundry apps deploy --change-type Patch --change-log "msg" --no-prompt  # Deploy to cloud
foundry apps release                                             # Release to catalog

Troubleshooting

Skills not invoked

If Claude doesn't use Foundry skills automatically, phrase your prompt with a clear action verb and Foundry noun (e.g., "create a foundry app", "fix the foundry function"). You can also say "Use Foundry skills" at any point to redirect.

CLI authentication

foundry profile active    # Check current profile
foundry login             # Re-authenticate
foundry profile list      # List all profiles

Deployment failures

1. Validate immediately after foundry api-integrations create (foundry apps validate --no-prompt) — Foundry's server-side OpenAPI parser is stricter than redocly lint and may reject large vendor specs
2. Run foundry apps deploy from the project root directory
3. Check the debugging-workflows skill for systematic troubleshooting

Testing

Three scripts validate skill changes at different levels. All require macOS or Linux (bash).

Tip: Wrap long-running tests with caffeinate -i to prevent macOS from sleeping mid-run:

caffeinate -i ./run-ab-test.sh --fresh 5

Hook tests

./test-hooks.sh

Unit tests for the four hook scripts (skill router, superpowers bridge, CLI guard, session start). Fast, no API calls, no Foundry CLI needed. Run after any hook change.

Skill test (single run)

./test-skill.sh                    # 5 runs against local plugin
./test-skill.sh --runs 1           # Quick single run
./test-skill.sh --plugin-dir .     # Explicit plugin path (default is ".")

Runs the example Okta prompt end-to-end: scaffolds an app, deploys it, and scores the result. Each run takes 5-10 minutes and costs ~5-10M tokens. Results go to /tmp/foundry-skill-test/.

Verify apps (after test-skill)

SKIP_BROWSER=1 ./verify-apps.sh                           # Phase 1 only: spec analysis + release
OKTA_DOMAIN=... OKTA_API_KEY=... ./verify-apps.sh          # Phase 1 + Phase 2: browser install/uninstall

Verifies apps created by test-skill.sh or A/B tests — analyzes OpenAPI specs, releases apps, and optionally installs them in the Falcon console via browser automation.

./verify-apps.sh                   # Verify test-skill.sh runs (default: /tmp/foundry-skill-test/)
./verify-apps.sh --green           # Verify GREEN phase of A/B test
./verify-apps.sh --dir /path       # Verify runs in a specific directory

A/B test (main vs local branch)

./run-ab-test.sh                   # 5 runs per phase (baseline: main)
./run-ab-test.sh 1                 # Quick single run per phase
./run-ab-test.sh 5 --ref v1.0.0    # Compare against a specific release tag
./run-ab-test.sh --no-skill        # No plugin vs local plugin (1 run each)

To compare against the last release with a fresh baseline, without macOS sleeping mid-run:

caffeinate -i ./run-ab-test.sh --ref "$(git describe --tags --abbrev=0)" --fresh 5

Compares baseline ref skills (RED) against local branch skills (GREEN). The --ref flag accepts any git ref (tag, branch, commit SHA) and defaults to main. Results saved to /tmp/foundry-skill-ab/baseline.json and /tmp/foundry-skill-ab/optimized.json. Tokens and cost come from each session's final result event, so compare cost per run first; a baseline saved before that change is re-run automatically.

Contributing

The skills improve every time someone uses them to build an app. If you hit a rough edge or find that Claude struggles with a particular pattern, you can teach the skills to handle it better.

Use the skills, then improve them

1. Clone this repo and configure your AI coding assistant (see Getting Started)
2. Try a multi-capability prompt (see Example prompt above)
3. Watch for patterns where Claude struggles, retries, or produces incorrect output
4. At the end of the session, ask Claude to fix the skills directly:

What did you learn from this session that could improve the Foundry skills?
Clone https://github.com/CrowdStrike/foundry-skills.git,
create a branch, update the skills with this knowledge, and
create a PR on GitHub.

Claude handles the branch, skill edits, and PR creation. Even when Claude struggles to build an app, it usually figures it out eventually. This step captures that learning so the next session is faster and uses fewer tokens.

Development workflow

1. Clone the repo (see Getting Started)
2. Edit skill files in skills/*/SKILL.md
3. Run ./test-hooks.sh to validate hooks
4. Test with ./test-skill.sh --runs 1 for a quick end-to-end check
5. Run ./run-ab-test.sh 1 to compare against main before opening a PR

Release process

Releasing these skills is easy! Just run the following command:

./release.sh

This walks you through a semantic version bump (major/minor/patch), updates the version in plugin.json, marketplace.json, README badge, all SKILL.md files, and the CHANGELOG date, then creates a release branch and PR. After the PR is approved and merged, create a draft GitHub release to tag main:

gh release create v<version> --target main --title "v<version>" --generate-notes --draft

This generates release notes from merged PRs and saves them as a draft. Review and edit the notes at github.com/CrowdStrike/foundry-skills/releases, then click Publish when ready.

After publishing the release, update the marketplaces that pin a snapshot. Build and upload the OpenAI skills-only bundle with ./release.sh --bundle v<version>, email the Cursor marketplace team to re-index the listing, and open a PR to github/awesome-copilot that bumps the version, ref, and sha of the crowdstrike-falcon-foundry entry. release.sh prints the exact steps when it finishes.

Cross-Plugin: Fusion Workflows

foundry-skills builds Falcon Foundry apps with the full lifecycle: API integrations, UI pages, serverless functions, collections, and workflows wrapped in a manifest.yml. When a request is a standalone Fusion workflow (a trigger plus actions that already exist, with no app to build), use the sibling plugin (fusion-skills):

/plugin install crowdstrike-falcon-fusion

The two plugins detect each other and advise the right path. Use foundry-skills for the full app lifecycle; use fusion-skills for standalone workflows with live action discovery. Neither requires the other to function.

License

See LICENSE for details.