Skip to main content
Give your AI coding agent what it needs to run c1i safely against your C1 tenant. The c1i docs agents command exports a self-contained skill file that teaches the agent c1i’s conventions, and the agent discovers individual commands and flags from c1i’s built-in help.

What is a skill file?

A skill file is a structured Markdown document that gives an AI coding agent the context it needs to use a tool effectively. The skill file produced by c1i docs agents covers:
  • Choosing the right tenant and authenticating, including limiting the credential to specific roles
  • Output formats, field selection, and exit codes
  • Choosing between a first-class command and a raw API call
  • Finding API endpoints, documentation pages, and the built-in runbooks (c1i docs guide)
  • Pagination, dry runs, and behavior that commonly surprises agents
  • Upgrading c1i
The skill file doesn’t list every command and flag. The agent reads those from c1i --help and c1i <command> --help, which always match the installed version. When you add this skill file to your agent’s context, the agent can manage users, apps, entitlements, tasks, and access requests in C1 through natural-language instructions.

Generate the skill file

Run c1i docs agents to output the skill file to your terminal:
To write it directly to a file:
The docs agents command works without authentication, so you can generate the skill file before logging in. c1i docs skill is an alias that prints the same file.

Set up c1i with Claude Code

Claude Code loads skills from .claude/skills/ in your project directory. To add c1i as a skill:
1
Create the skills directory if it doesn’t exist:
2
Generate the skill file into that directory:
3
Verify the file was created:
You should see the skill content starting with a YAML frontmatter block containing name: c1i.
4
Start using c1i through Claude Code. You can now give Claude natural-language instructions like:
  • “List all users with status enabled”
  • “Search for entitlements related to admin access”
  • “Show me all open access request tasks”
  • “Create a grant request for user X to entitlement Y”
Make sure c1i is installed and authenticated (c1i auth login) before asking Claude Code to run c1i commands. To give the agent less than your full access, log in with --scoped-role or --choose-roles; see Limit the credential to specific roles. For non-interactive setups (CI, containers, headless servers), set C1I_URL, C1I_CLIENT_ID, and C1I_CLIENT_SECRET as environment variables instead of running auth login — c1i picks them up automatically. See Credential storage for the full precedence order.

Set up c1i with Cursor

Cursor uses project rules to provide context to its AI. To add the c1i skill:
These examples use conductor.one. If your organization is on the EU data residency instance, substitute c1eu.ai in URLs, client IDs, and hostnames.

Set up c1i with other agents

The skill file is plain Markdown with YAML frontmatter, so it works with any AI coding agent that accepts context files. The general pattern:
1
Generate the skill file:
2
Add the file wherever your agent reads context. Common locations:
3
Authenticate c1i if you haven’t already:
Your agent now reads the c1i skill from its context and can run c1i commands against your tenant.

Example c1i agent workflows

Once your agent has the c1i skill, you can use natural language to accomplish common tasks.

Confirm the active identity

Ask your agent to verify which user it’s acting as before making changes:
The agent runs c1i auth whoami and reports the user ID, display name, email, counts of roles, permissions, and feature flags, and the tenant URL it’s targeting.

Explore the API

Ask your agent to discover what’s available before running authenticated commands:
The agent uses c1i docs endpoints and c1i docs endpoint to explore the API without credentials.

Audit user access

The agent runs c1i users list --status=enabled and c1i entitlements list --app-id=<id> and presents the results. For broader queries where you want a fast answer rather than a full dataset, ask the agent to limit results:
The agent runs c1i users list --status=enabled --limit=5. c1i tightens the per-call request size to match --limit and stops auto-paginating once the cap is reached, so the API call is small even when the underlying tenant has thousands of users.

Manage access requests

The agent runs c1i tasks list --state=open --assigned-to-me, filters the results, and uses c1i tasks approve for the matching tasks.

Investigate an app’s accounts

The agent runs c1i accounts list --app-id=<id> --unmapped-only to surface accounts that haven’t been linked to C1 users.

Keep the skill file up to date

The skill file includes the c1i version in its frontmatter. When you update c1i, regenerate the skill file to pick up new commands and changes: