didit-protocol/skills

didit/skills Official AI Agent Skills for the Didit Identity Verification Platform

didit.me · API Docs · Business Console · Agent Skills Standard

---

20 production-ready skills that teach AI agents how to call the Didit API: account registration, identity verification, KYC workflows, AML screening, biometric checks, fraud checks, transaction and event monitoring, cases, fraud networks, SDK integration, billing and user management. Works with Cursor, Claude Code, Codex, OpenCode, and any agent that supports the Agent Skills standard.

What’s Inside

Naming convention: Domain-term names that match how users naturally ask: face-match, proof-of-address, aml-screening, kyc-onboarding. Lowercase kebab-case, didit- namespaced. Every skill that uses an API key includes a Getting Started section for account creation and credits.

For any of the seven product combinations, start with the three-product integration prompt. It keeps credentials and decisions separate and links the person through server-owned vendor_data. Authentication SDKs, JWT/JWKS, automatic auth events and Authentication step-ups are release-dependent. Dedicated Authentication, electronic-signature and biometric-consent skills await merged, documented API contracts; they are not part of the 20-skill catalog yet.

---

Zero to Verifying (No Browser Needed)

Agents can go from nothing to a live verification link in 5 API calls:

1. POST /programmatic/register/           → register with your email
2. POST /programmatic/verify-email/        → verify the emailed code, get application.api_key
3. POST /v3/workflows/                     → create a workflow (features array)
4. POST /v3/webhook/destinations/          → register a webhook URL and the events it receives
5. POST /v3/session/                       → create a session → get the verification URL

---

Install

Option 1: npx skills (recommended)

Install all skills at once:

npx skills add didit-protocol/skills

Or install a specific skill:

npx skills add didit-protocol/skills --skill didit-face-match

Works with Cursor, Claude Code, Codex, and OpenCode. The CLI auto-detects your agent and installs to the correct directory.

Option 2: Git Clone

git clone https://github.com/didit-protocol/skills.git

Copy the skills you need from skills/ into your agent’s skill directory.

Option 3: Manual

Copy the entire skills// folder, including SKILL.md and scripts/, into your agent’s skill directory. The helpers referenced by a skill must travel with it.

---

Setup

Option A: Programmatic Registration (recommended for agents)

No console needed: agents can register themselves.

# 1. Register
curl -X POST https://apx.didit.me/auth/v2/programmatic/register/ \
  -H "Content-Type: application/json" \
  -d '{"email": "[email protected]", "password": "MyStr0ng!Pass"}'

# 2. Check your email for the 6-character code, then verify
curl -X POST https://apx.didit.me/auth/v2/programmatic/verify-email/ \
  -H "Content-Type: application/json" \
  -d '{"email": "[email protected]", "code": "A3K9F2"}'
# → the API key is in the response at application.api_key

export DIDIT_API_KEY="the_api_key_from_response"

Option B: Business Console

1. Get an API key in the Didit Business Console under Developers > API keys. Use the primary secret or create a named key with only the access the agent needs.
2. Set the environment variable:

export DIDIT_API_KEY="your_api_key"

3. (Optional) For webhook verification, keep the destination’s signing secret (secret_shared_key, returned when you create the destination and by GET /v3/webhook/destinations/{id}/):

export DIDIT_WEBHOOK_SECRET="your_secret"

4. Talk to your agent. Once skills are installed, just ask naturally:

“Register a Didit account for me” “Create a KYC workflow with ID scan and liveness” “Create a verification session for this user” “Check my credit balance” “Screen John Smith against AML databases” “Add the Didit web SDK to my checkout page”

Option C: MCP server

To let an AI assistant work in your Didit account as you (no API key), connect the hosted MCP server at https://mcp.didit.me/mcp. See the didit-mcp-server skill.

---

API Endpoints Covered

130 endpoint paths across the platform, most with several methods.

---

Testing

Offline checks (format, README, scripts, publication filters, code examples, webhook signatures). Python and Git are required; no key is needed:

python3 tests/test_skill_structure.py

Live checks against the real API (a sandbox application key is recommended; some calls are billed on live applications):

export DIDIT_API_KEY="your_key"
export DIDIT_WORKFLOW_ID="your_workflow_id"     # optional
export DIDIT_ORG_ID="your_organization_id"      # optional, enables case and network checks
export DIDIT_APP_ID="your_application_id"       # optional, enables case and network checks
python3 tests/test_all_skills.py

Each live check passes only when the endpoint answers with the documented status code and JSON fields.

---

Repo Structure

skills/                              ← 20 skills (1 hub + 1 KYC recipe + 18 product skills)
├── didit-verification-management/    SKILL.md + scripts/{setup_account,manage_workflows,create_session}.py
├── didit-kyc-onboarding/             SKILL.md + scripts/run_kyc.py
├── didit-id-document-verification/   SKILL.md + scripts/verify_id.py
├── didit-liveness-detection/         SKILL.md + scripts/check_liveness.py
├── didit-face-match/                 SKILL.md + scripts/match_faces.py
├── didit-face-search/                SKILL.md + scripts/search_faces.py
├── didit-biometric-age-estimation/   SKILL.md + scripts/estimate_age.py
├── didit-email-verification/         SKILL.md + scripts/verify_email.py
├── didit-phone-verification/         SKILL.md + scripts/verify_phone.py
├── didit-aml-screening/              SKILL.md + scripts/screen_aml.py
├── didit-proof-of-address/           SKILL.md + scripts/verify_address.py
├── didit-database-validation/        SKILL.md + scripts/validate_database.py
├── didit-document-ai/                SKILL.md + scripts/extract_document.py
├── didit-transaction-monitoring/     SKILL.md + scripts/submit_transaction.py
├── didit-event-monitoring/           SKILL.md + scripts/send_event.py
├── didit-fraud-check/                SKILL.md + scripts/run_fraud_check.py
├── didit-case-management/            SKILL.md + scripts/manage_cases.py
├── didit-fraud-networks/             SKILL.md + scripts/list_networks.py
├── didit-sdk-integration/            SKILL.md + scripts/mint_sdk_tokens.py
└── didit-mcp-server/                 SKILL.md
tests/test_skill_structure.py        ← offline checks
tests/test_all_skills.py             ← live endpoint checks

Each SKILL.md follows the three-tier information architecture:

1. Metadata (always loaded): Domain-term name + trigger-based description in YAML frontmatter (~100 tokens)
2. Skill body (loaded on invocation): Full API documentation, examples, and workflows
3. Scripts (lazy loaded): Python CLI utilities referenced by the skill body

---
name: didit-verification-management
description: >
  Full Didit identity verification platform management...
version: 5.1.0
metadata:
  openclaw:
    requires:
      env: []
    primaryEnv: DIDIT_API_KEY
    emoji: "🛡️"
    homepage: https://docs.didit.me
---

---

Contributing

1. Fork the repo
2. Update or add a skill in skills/
3. Run python3 tests/test_skill_structure.py, and python3 tests/test_all_skills.py with a sandbox key
4. Open a PR

---

ClawHub publication handoff

Publication is a human step. The seven SK1 additions below still need ClawHub publication; this repository change prepares their bundles and does not publish them.

Before publishing, check the registry’s existing version for each skill and choose an unused version if necessary. Previously published skills affected by the script exclusion also need corrected bundles.

Stage each complete skill folder in a clean temporary directory. Include the repository .gitignore and .clawhubignore at that folder’s root, appending any skill-local rules to the matching policy file. ClawHub reads these files from the publication folder; it does not inherit the repository’s filters when publishing a nested skill directly. Keep scripts/ included and exclude credentials, caches and development files. See the ClawHub skill format.

Run the offline checks above, inspect the staged file list and confirm every referenced helper is present. Publish each staged skill folder individually, using its name and reviewed version, only when the human publisher is ready. After publication, install into an empty directory and verify SKILL.md, the helper files and their --help commands before announcing availability.

---

License

MIT, see LICENSE.