Kessel Onboarding Agents

AI agents and skills for the Kessel service onboarding program. Jira remains the system of record; agents propose artifacts and humans approve writes.

Quick start

New here? Start with the interview. Jira access is only needed if you plan to provision issues — the interview and schema design skills work without it.

Install the plugin (see Installing this plugin below).

Run an intake session:

/kessel-onboarding:interview --provider "Subscription Management" --service "Activation Keys"

Review the generated ServiceProfile and handoff in ./artifacts/profiles/.

Optionally design schemas:

/kessel-onboarding:schema-design --profile artifacts/profiles/{slug}-profile.json \
  --codebase_ref ~/dev/my-service

To preview what would be created in Jira (no writes):

/kessel-onboarding:provision --handoff artifacts/profiles/{slug}-handoff.md --dry-run

The quick-start examples use Claude Code syntax. For OpenAI substitutions, see Command examples.

Only need steps 2–3? Skip preflight. It is only required before live Jira provisioning (provision without --dry-run).

Installing this plugin

Claude Code marketplace (recommended for Claude Code)

The plugin is hosted as a Claude Code marketplace on GitHub. Install it permanently in one step:

# 1. Register the Kessel marketplace (one-time)
/plugin marketplace add project-kessel/kessel-onboarding-agents

# 2. Install the plugin from that marketplace
/plugin install kessel-onboarding@kessel-onboarding-agents

Once installed, all slash commands are available in every Claude Code session without any flags.

Codex CLI

Install the plugin from the GitHub marketplace with Codex CLI:

codex plugin marketplace add https://github.com/project-kessel/kessel-onboarding-agents
codex plugin add kessel-onboarding@kessel-onboarding-agents

Start a new Codex session after installing the plugin. Codex's native marketplace format normally uses .agents/plugins/marketplace.json, but Codex CLI accepts this repository's existing .claude-plugin/marketplace.json as a legacy-compatible marketplace and converts .claude-plugin/plugin.json during installation. No second marketplace manifest is required here.

Note

The same Github URL can be used for adding the plugin in ChatGPT Desktop UI by navigating to "Plugins" --> "Add" --> "Add a Marketplace"

Claude Code (CLI or desktop app) — local or session-only

# Session-only — available for this session only, nothing written to global config
claude --plugin-dir /path/to/kessel-onboarding-agents

# Persistent local install — inside a Claude Code session
/install-plugin /path/to/kessel-onboarding-agents

Once loaded, all slash commands listed below become available in Claude Code.

Cursor

Cursor uses a different skill system from Claude Code. Skills are SKILL.md files placed in:

User-level (available in all projects): ~/.cursor/skills/{skill-name}/SKILL.md
Project-level (available only in the current repo): .cursor/skills/{skill-name}/SKILL.md

To make the kessel-onboarding skills available in Cursor, copy or symlink the skill directories you want into one of those locations:

# User-level — available across all your Cursor projects
mkdir -p ~/.cursor/skills
cp -r /path/to/kessel-onboarding-agents/skills/onboarding-schema-design ~/.cursor/skills/
cp -r /path/to/kessel-onboarding-agents/skills/onboarding-validate-interview ~/.cursor/skills/

# Or project-level — only in this repo
mkdir -p .cursor/skills
cp -r /path/to/kessel-onboarding-agents/skills/onboarding-schema-design .cursor/skills/

Cursor does not use the agents/ or commands/ directories — only the skills/*/SKILL.md files. The slash command interface (/kessel-onboarding:interview) is Claude Code-specific; in Cursor, reference the skill by asking the agent to follow the instructions in the SKILL.md file, or by including it as a Cursor rule.

The --plugin-dir flag is Claude Code CLI only — Cursor does not support it.

Setting up for Jira provisioning

The interview and schema-design skills work without any Jira configuration. To enable live Jira provisioning (/kessel-onboarding:provision without --dry-run):

Create ~/.config/kessel-onboarding/config.json with the following required keys:
```
{
  "jira_host": "redhat.atlassian.net",
  "jira_cloud_id": "YOUR_CLOUD_ID",
  "initiative_project": "YOUR_INITIATIVE_PROJECT",
  "onboarding_project": "YOUR_ONBOARDING_PROJECT",
  "onboarding_label": "kessel-onboarding",
  "platform_gates_path": "context/platform-gates.json",
  "artifacts_dir": "./artifacts",
  "mcp_server_name": "YOUR_ATLASSIAN_MCP_SERVER_NAME",
  "team_field_id": "YOUR_TEAM_FIELD_ID",
  "team_field_value": "YOUR_TEAM_UUID"
}
```
- initiative_project — the Jira project where Provider Initiatives are created (e.g. your team's feature/initiative tracking project)
- onboarding_project — the Jira project where Service Epics and Phase Stories are created (e.g. your team's sprint/work-item project)
- team_field_value — the UUID of your team's entry in the Jira "Team" field (not the display name).
See docs/configuration.md for all field descriptions, how to find your jira_cloud_id, and how to look up your team UUID.
Create a .env file with your Atlassian API credentials (see docs/configuration.md for the REST fallback setup).
Run /kessel-onboarding:preflight --provisioner to validate all checks required for live provisioning.

Recommended models

Skill complexity varies — some tasks are structured Q&A, others generate KSL and authorization code where errors have downstream consequences. In our experience, relying on Sonnet for Claude (or if you have the budget, OpusPlan is better) and Luna xhigh/max for Codex has been sufficient for the entire suite of skills, especially if you are trying to remain cost conscious.

For more complex code bases or scenarios, some skills may benefit from higher reasoning/context models. Should you desire to run specific skills using different models, below provides a reasonable list of recommended models for each skill process.

Claude Code

Skill	Recommended model
`migrate-rbac-v1`, `validate-interview`	Opus/OpusPlan: code generation and multi-repo scoring need the strongest reasoning
`schema-design`, `suggest-patterns`	OpusPlan or Sonnet acceptable for simple services
`interview-conduct`	Sonnet - structured Q&A with confirm-or-correct drafting
`dedup-epic`, `preflight`	Sonnet - mechanical JQL and connectivity checks

The /model opusplan setting (Opus for planning turns, Sonnet for execution) is a reasonable default for most sessions.

Codex (GPT-5.6 family)

Skill	Recommended Model
`migrate-rbac-v1`, `validate-interview`	Sol/Terra high
`schema-design`, `suggest-patterns`	Terra high or Luna Max
`interview-conduct`	Terra medium or Luna xhigh/max
`dedup-epic`, `preflight`	Luna high

Sol is the most capable but expensive — reserve it for the two skills where a wrong output (incorrect KSL, missed validation dimension) causes real rework only.

Skills and agents

Full onboarding pipeline

                    ┌── provision          (creates Jira tracking issues)
preflight → interview ─┤
                    └── schema-design → migrate-rbac-v1   (implementation)

validate-interview  (run against already-migrated services, independent)

Skill reference

Codex invocation syntax

Codex does not use Claude Code's /plugin:skill syntax. For workflows that chain multiple skills, invoke the plugin first and then the namespaced workflow:

@kessel-onboarding onboarding:interview --provider "UIE" --service "Task Manager" --codebase_ref ~/dev/task-manager

.

Command	What it does	When to use
`/kessel-onboarding:preflight`	Validates config, MCP connection, Jira access, and JQL templates	Before live provisioning on a new machine
`/kessel-onboarding:interview`	Phase 0/1 intake — structured Q&A with EM/tech lead, builds ServiceProfile, suggests adoption patterns, dedups Jira	Start of every new service onboarding
`/kessel-onboarding:schema-design`	Phase 2 head start — analyzes service codebase and generates draft resource schemas (inventory-api) and permissions schemas (rbac-config KSL + JSON)	After the interview, before Phase 2 begins
`/kessel-onboarding:provision`	Dry-run then create Jira issue batch (Initiative + Epic + Phase Stories) from an approved handoff	After the interview, when you're ready to provision Jira issues
`/kessel-onboarding:validate-interview`	Scores an interview's accuracy by comparing its ServiceProfile against the service's actual Kessel implementation	After Phase 4+, or retroactively on already-migrated services
`/kessel-onboarding:migrate-rbac-v1`	Finds v1 RBAC call sites in the service codebase, classifies each by KSL-016 pattern, and writes Kessel v2 replacement code (uncommitted)	After schema-design — translates onboarding decisions into actual code changes
`/kessel-onboarding:test`	Full test loop — interview + schema-design with Kessel blindfold on, then auto-validate against real implementation	Against any already-migrated service; primary tool for measuring skill accuracy

Note

For OpenAI tools, use the plugin-qualified form for chained workflows. Targeting an individual skill directly with @plugin-name or $skill-name does not reliably preserve the handoff between skill steps. The Codex form maps the Claude Code command /kessel-onboarding:interview to @kessel-onboarding onboarding:interview, flags can still be provided as expected, OpenAI models will interpret them correctly. See Command Examples below for exact commands

Agents

Agent	Orchestrates	Command
Onboarding Interview	`onboarding-interview-conduct` → `onboarding-interview-suggest-patterns` → `onboarding-dedup-epic` → `onboarding-format-handoff`	`/kessel-onboarding:interview`
Onboarding Provisioner	`onboarding-provision-jira`	`/kessel-onboarding:provision`

The remaining skills (onboarding-schema-design, onboarding-validate-interview, onboarding-migrate-rbac-v1, onboarding-preflight) are standalone — they do not require a full agent and can be invoked directly via their slash commands.

When to run each skill

New service → /kessel-onboarding:preflight           (once per machine, only needed for live Jira provisioning)
           → /kessel-onboarding:interview            (Phase 0/1 intake — produces profile + handoff)
           ├── /kessel-onboarding:provision          (creates Jira tracking issues from the handoff)
           └── /kessel-onboarding:schema-design      (Phase 2 — generates KSL + resource schemas)
                   └── /kessel-onboarding:migrate-rbac-v1  (Phase 4 — writes v1→v2 code changes)

Already-migrated service
           → /kessel-onboarding:validate-interview  (score the interview vs real implementation)
           → /kessel-onboarding:test               (full automated loop: blindfolded run + auto-validate)

schema-design requires codebase access and will ask for a repo URL or local path if the interview did not capture one. It generates local draft files only — no PRs are opened.

validate-interview is most useful after Phase 4 (PoC) when the service's Kessel integration is established. Running it retroactively on services like HBI gives a baseline for how accurately the interview predicts implementation needs.

Command examples

The examples below use Claude Code's slash-command syntax. In Codex, use @kessel-onboarding onboarding:<skill-name> with the same flags.

`/kessel-onboarding:preflight`

# Run before first use on a machine, or after any config change
/kessel-onboarding:preflight
# or For OpenAI
@kessel-onboarding onboarding:preflight

`/kessel-onboarding:interview`

For OpenAI, substitute /kessel-onboarding:interview with @kessel-onboarding onboarding:interview for any commands below

# Simplest — just the service name; Claude will ask for everything else
/kessel-onboarding:interview "Activation Keys"

# For OpenAI
@kessel-onboarding onboarding:interview

# With provider context (avoids being asked for it during the session)
/kessel-onboarding:interview --provider "Subscription Management" --service "Activation Keys"

# With a known feature epic to link
/kessel-onboarding:interview --provider "Subscription Management" --service "Activation Keys" \
  --feature-epic TUSC-271

# Headless — build profile from an existing notes file, skip live Q&A
/kessel-onboarding:interview --service "Activation Keys" \
  --headless examples/activation-keys-intake-notes.md

# Save only — run the interview but don't offer provisioner dispatch at the end
/kessel-onboarding:interview --provider "Insights" --service "Host Based Inventory" \
  --save-only

# Test mode — Kessel blindfold on; artifacts go to artifacts/test/{slug}/ (for skill testing)
/kessel-onboarding:interview --provider "Insights" --service "Host Based Inventory" \
  --test-mode

`/kessel-onboarding:schema-design`

For OpenAI, substitute /kessel-onboarding:schema-design with @kessel-onboarding onboarding:schema-design for any commands below

# Simplest — profile only; Claude will ask for codebase reference before continuing
/kessel-onboarding:schema-design \
  --profile artifacts/profiles/activation-keys-profile.json

# With codebase reference (avoids being prompted)
/kessel-onboarding:schema-design \
  --profile artifacts/profiles/activation-keys-profile.json \
  --codebase_ref path/to/codebase

# Custom output directory
/kessel-onboarding:schema-design \
  --profile artifacts/profiles/activation-keys-profile.json \
  --codebase_ref path/to/codebase \
  --output_dir ~/scratch/activation-keys-schemas

# Test mode — ignores existing KSL and inventory-api schemas; outputs to artifacts/test/{slug}/schemas/
/kessel-onboarding:schema-design \
  --profile artifacts/test/host-based-inventory/profiles/host-based-inventory-profile.json \
  --codebase_ref path/to/codebase \
  --test-mode

`/kessel-onboarding:provision`

For OpenAI, substitute /kessel-onboarding:provision with @kessel-onboarding onboarding:provision for any commands below

# Simplest — always dry-runs first and waits for approval before creating anything
/kessel-onboarding:provision \
  --handoff artifacts/profiles/activation-keys-handoff.md

# Explicit dry-run only — present the proposed issue table, stop before the approval prompt
/kessel-onboarding:provision \
  --handoff artifacts/profiles/activation-keys-handoff.md \
  --dry-run

# Skip the dry-run prompt if you've already reviewed the table
/kessel-onboarding:provision \
  --handoff artifacts/profiles/activation-keys-handoff.md \
  --confirm

`/kessel-onboarding:validate-interview`

For OpenAI, substitute /kessel-onboarding:validate-interview with @kessel-onboarding onboarding:validate-interview for any commands below

# Simplest — codebase only; skips KSL and resource schema dimensions
/kessel-onboarding:validate-interview \
  --profile artifacts/profiles/host-based-inventory-profile.json \
  --codebase_ref path/to/codebase

# With rbac-config — adds KSL, permissions.json, and roles.json scoring
/kessel-onboarding:validate-interview \
  --profile artifacts/profiles/host-based-inventory-profile.json \
  --codebase_ref path/to/codebase \
  --rbac_config_path ~/dev/rbac-config

# Full — all repos; scores all 12 interview dimensions
/kessel-onboarding:validate-interview \
  --profile artifacts/profiles/host-based-inventory-profile.json \
  --codebase_ref path/to/codebase \
  --rbac_config_path ~/dev/rbac-config \
  --inventory_api_path path/to/cloned-inventory-api-repo

# Full + schema artifacts — adds 4 schema-design dimensions (dimensions 13–16)
/kessel-onboarding:validate-interview \
  --profile artifacts/profiles/host-based-inventory-profile.json \
  --codebase_ref path/to/codebase \
  --rbac_config_path ~/dev/rbac-config \
  --inventory_api_path path/to/cloned-inventory-api-repo \
  --schema_artifacts_path artifacts/schemas/host-based-inventory

`/kessel-onboarding:migrate-rbac-v1`

For OpenAI, substitute /kessel-onboarding:migrate-rbac-v1 with @kessel-onboarding onboarding:migrate-rbac-v1 for any commands below

# Best path — after schema-design deferred migration (context has v2 names + patterns pre-loaded)
/kessel-onboarding:migrate-rbac-v1 \
  --rbac /path/to/insights-rbac \
  --context artifacts/schemas/task-manager/migrate-context.md \
  ~/dev/my-service

# After an interview only — use the profile to skip pattern re-derivation
/kessel-onboarding:migrate-rbac-v1 \
  --rbac /path/to/insights-rbac \
  --rbac-config /path/to/rbac-config \
  --profile artifacts/profiles/task-manager-profile.json \
  ~/dev/my-service

# Standalone — no prior onboarding, discovers everything from the repo
/kessel-onboarding:migrate-rbac-v1 \
  --rbac /path/to/insights-rbac \
  --rbac-config /path/to/rbac-config \
  ~/dev/my-service

# With inventory-api (needed for native / native-ws-list resource schemas)
/kessel-onboarding:migrate-rbac-v1 \
  --rbac /path/to/insights-rbac \
  --rbac-config /path/to/rbac-config \
  --inventory-api /path/to/inventory-api \
  --context artifacts/schemas/task-manager/migrate-context.md \
  ~/dev/my-service

`/kessel-onboarding:test`

For OpenAI, substitute /kessel-onboarding:test with @kessel-onboarding onboarding:test for any commands below

# Simplest — interview + schema-design (blindfolded) + validate; Claude finds rbac-config/inventory-api automatically
/kessel-onboarding:test \
  --service "Host Based Inventory" \
  --codebase_ref path/to/codebase

# With provider context
/kessel-onboarding:test \
  --service "Host Based Inventory" \
  --provider "Insights" \
  --codebase_ref path/to/codebase

# Full — explicit repo paths for complete 16-dimension scoring
/kessel-onboarding:test \
  --service "Host Based Inventory" \
  --provider "Insights" \
  --codebase_ref path/to/codebase \
  --rbac_config_path ~/dev/rbac-config \
  --inventory_api_path path/to/cloned-inventory-api-repo

Program docs (canonical)

Jira

Filter (all onboarding): labels = kessel-onboarding
Completed services: labels = kessel-onboarding AND type = Epic AND status = Done
Office hours: Tuesdays 8:00–8:30am ET — meet.google.com/exy-ohre-yov

Support: #forum-mgmt-fabric

Changelog

2026-09: Added OpenAI/Codex installation and invocation guidance throughout the README, including migrate-rbac-v1 command examples.
2026-07: Added test command and --test-mode flag to interview and schema-design; added command examples section covering simple to fully-flagged invocations for all skills.
2026-07: Added schema-design and validate-interview skills; updated quick start, skill reference table, and loading instructions for Claude Code and Cursor.
2026-07: Reworded the platform-gates quick-start step — gate status now only affects pattern-suggestion confidence, since Jira gate-linking was removed (the linking skill never had a populated Jira key to link against).
2026-07: Corrected the support Slack channel and added /kessel-onboarding:preflight as quick-start step 3.

Assisted-by: Claude (Anthropic)

Name		Name	Last commit message	Last commit date
.claude-plugin		.claude-plugin
.github		.github
agents		agents
commands		commands
context		context
docs		docs
examples		examples
scripts		scripts
skills		skills
.coderabbit.yaml		.coderabbit.yaml
.gitignore		.gitignore
AGENTS.md		AGENTS.md
CLAUDE.md		CLAUDE.md
CONTRIBUTING.md		CONTRIBUTING.md
README.md		README.md

Navigation Menu

Repository navigation

About

Repository files navigation

Kessel Onboarding Agents

Quick start

Installing this plugin

Claude Code marketplace (recommended for Claude Code)

Codex CLI

Claude Code (CLI or desktop app) — local or session-only

Cursor

Setting up for Jira provisioning

Recommended models

Claude Code

Codex (GPT-5.6 family)

Skills and agents

Full onboarding pipeline

Skill reference

Codex invocation syntax

Agents

When to run each skill

Command examples

`/kessel-onboarding:preflight`

`/kessel-onboarding:interview`

`/kessel-onboarding:schema-design`

`/kessel-onboarding:provision`

`/kessel-onboarding:validate-interview`

`/kessel-onboarding:migrate-rbac-v1`

`/kessel-onboarding:test`

Program docs (canonical)

Jira

Changelog

About

Releases

Packages

Contributors

Languages

Latest commit

History

Folders and files

Repository files navigation

Kessel Onboarding Agents

Quick start

Installing this plugin

Claude Code marketplace (recommended for Claude Code)

Codex CLI

Claude Code (CLI or desktop app) — local or session-only

Cursor

Setting up for Jira provisioning

Recommended models

Claude Code

Codex (GPT-5.6 family)

Skills and agents

Full onboarding pipeline

Skill reference

Codex invocation syntax

Agents

When to run each skill

Command examples

/kessel-onboarding:preflight

/kessel-onboarding:interview

/kessel-onboarding:schema-design

/kessel-onboarding:provision

/kessel-onboarding:validate-interview

/kessel-onboarding:migrate-rbac-v1

/kessel-onboarding:test

Program docs (canonical)

Jira

Changelog

About

Resources

Contributing

Stars

Watchers

Forks

Releases

Packages

Contributors

Languages

`/kessel-onboarding:preflight`

`/kessel-onboarding:interview`

`/kessel-onboarding:schema-design`

`/kessel-onboarding:provision`

`/kessel-onboarding:validate-interview`

`/kessel-onboarding:migrate-rbac-v1`

`/kessel-onboarding:test`