Back to search
orgtomo
MarketplaceWorks with Claude CodeGitHub
A team of Claude Code agents that takes a Salesforce ticket from understanding to a verified deploy pack, with proof at every step.
Safety scan
A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.
Checked today
Badge for your README
Markdown
[](/listings/khadda-ai--orgtomo)HTML
<a href="/listings/khadda-ai--orgtomo"><img src="/api/listings/khadda-ai--orgtomo/badge.svg" alt="findsafeskills safety rating" /></a>Install
Install instructions from the README
As written by the repo's author. findsafeskills has not verified these steps.
npm install -g github:khadda-ai/orgtomo mkdir my-company && cd my-company orgtomo setup # asks for your orgs, deploy tools and allowed test data orgtomo doctor --orgs # checks the CLI, each org, and whether it is really a sandbox orgtomo brain sync uat # build the org map orgtomo discover run uat --retrieve # optional: study how the business uses the org orgtomo ui # review agents, safety and features, browse the org map
Install the three plugins for this folder only, so that they load here and nowhere else:
claude plugin marketplace add khadda-ai/orgtomo claude plugin install orgtomo@orgtomo --scope project claude plugin install salesforce-development@orgtomo --scope project claude plugin install salesforce-code-quality@orgtomo --scope project
Then start a session with the Lead and give it a ticket:
orgtomo start
/orgtomo:ping /orgtomo:ticket ABC-123
Full walkthrough: [docs/SETUP.md](docs/SETUP.md).
What was scanned
The files the safety scan read, as written by the repository's author and shown as plain text. Lines with a finding are marked.
.claude-plugin/marketplace.json · Manifest · 39 lines
- {
- "name": "orgtomo",
- "owner": {
- "name": "Khadda AI Labs"
- },
- "metadata": {
- "description": "Orgtomo and the Salesforce plugins it builds on. forcedotcom/sf-skills changes often; the two Salesforce plugins are pinned to one commit so that agent behaviour changes only when a maintainer moves the pin (see sf-skills.lock).",
- "version": "0.2.0"
- },
- "plugins": [
- {
- "name": "orgtomo",
- "description": "The Orgtomo team: commands, guards and the engine.",
- "source": "./"
- },
- {
- "name": "salesforce-development",
- "description": "Salesforce skills, LSP and metadata tools from forcedotcom/sf-skills, pinned to commit 788b83fb (plugin 2.8.0).",
- "version": "2.8.0",
- "source": {
- "source": "git-subdir",
- "url": "https://github.com/forcedotcom/sf-skills.git",
- "path": "plugins/builder/salesforce-development",
- "sha": "788b83fb9f5effb4f1548e3090df415704d3ec64"
- }
- },
- {
- "name": "salesforce-code-quality",
- "description": "Code Analyzer skills and the architecture review agent from forcedotcom/sf-skills, pinned to commit 788b83fb (plugin 0.2.0).",
- "version": "0.2.0",
- "source": {
- "source": "git-subdir",
- "url": "https://github.com/forcedotcom/sf-skills.git",
- "path": "plugins/builder/salesforce-code-quality",
- "sha": "788b83fb9f5effb4f1548e3090df415704d3ec64"
- }
- }
- ]
- }
README.md · README · 173 lines
- # Orgtomo
- A team of Claude Code agents that takes a Salesforce ticket from "what is being asked" to a verified deploy pack, with proof at every step.
- [](https://github.com/khadda-ai/orgtomo/actions/workflows/ci.yml)
- [](LICENSE)
- *Org* + *tomo* (友, Japanese for friend, companion).
- ## What it is
- Orgtomo runs inside [Claude Code](https://code.claude.com/docs). You give it a ticket key. A lead agent hands the work to specialists, one stage at a time, and an engine decides after each stage whether the result is good enough to continue.
- It is built for Salesforce developers, admins and consultants who already do this work with an AI assistant by hand: paste the ticket, ask for an explanation, ask for a reproduction, ask for a plan, build, test, write the update. Orgtomo makes that sequence a system, and adds the parts that are easy to skip when doing it by hand.
- **Nothing is guessed.** Every claim an agent makes must cite evidence that the engine itself recorded: a query result, a test run, a deploy, a fresh fact from the org map, or a sentence from official Salesforce documentation. Wording such as "probably" fails a gate.
- **The Lead does not trust reports.** After a stage passes its gates, the Lead re-runs the proof of one claim and compares.
- **Production is read-only.** Reads are masked. There is no setting that changes this.
- **Humans deploy.** Agents build and test in your dev sandbox. For every other org, Orgtomo prepares a deploy pack for your tool (Blue Canvas, change sets, Git and the CLI, or a generic checklist), waits for you, then checks that the org matches what was built.
- **Test data is safe.** Only the e-mail addresses and phone numbers you allow may appear in test records, scripts, browser input and drafts. In the default mode, agents write data only while the sandbox cannot send e-mail.
- **It learns from what really happened.** Failed gates, rejected plans, reviewer findings and your own feedback are recorded as observations. Lessons must cite them. Approved lessons are loaded by the agent they concern.
- ## The team
- | Agent | Role |
- |---|---|
- | Tomo Lead | Architect and manager. Delegates, verifies, reports. Never builds. |
- | Analyst | Business analyst, functional consultant and product owner: plain and technical understanding, acceptance criteria, examples. |
- | Reproducer | Proves a bug with data in a sandbox. Never by reading code. |
- | Planner | Smallest correct design, grounded in the org map and official documentation. |
- | Builder | Implements the approved plan in the build org: backup first, your naming and comment style. |
- | Tester | Tests every criterion with proof, reruns the reproduction, bulk and regression. |
- | Reviewer | Independent review against a rubric, on a different model from the Builder. |
- | Cartographer, Baseline, Scribe, Coach, Notetaker | Org map, before-state, message drafts, lessons, meeting notes. Optional ones have a switch. |
- ## How a ticket moves
- ```mermaid
- flowchart LR
- open --> understand --> repro --> plan --> build --> test --> review --> handoff --> close --> done
- test -. "fails: back to plan (max 2 loops, then stop)" .-> plan
- review -. "changes requested" .-> plan
- plan -- "human approves" --> build
- review -- "human approves" --> handoff
- handoff -- "human deploys, per org" --> handoff
- ```
- Stages, gates and contracts are data (`spec/stages.yaml`, `spec/gates.yaml`, `spec/schemas/`). The engine interprets them.
- ## Requirements
- - Claude Code
- - Node.js 20 or newer
- - Salesforce CLI (`sf`), logged in to each org under an alias
- - Optional: [Obsidian](https://obsidian.md) to browse the vault (the control panel shows the org map too); a tracker connector (Atlassian by default); the Playwright MCP server for browser steps
- Orgtomo has no runtime dependencies to install. The one third-party library it uses is vendored.
- ## Quick start
- ```bash
- npm install -g github:khadda-ai/orgtomo
- mkdir my-company && cd my-company
- orgtomo setup # asks for your orgs, deploy tools and allowed test data
- orgtomo doctor --orgs # checks the CLI, each org, and whether it is really a sandbox
- orgtomo brain sync uat # build the org map
- orgtomo discover run uat --retrieve # optional: study how the business uses the org
- orgtomo ui # review agents, safety and features, browse the org map
- ```
- Install the three plugins for this folder only, so that they load here and nowhere else:
- ```bash
- claude plugin marketplace add khadda-ai/orgtomo
- claude plugin install orgtomo@orgtomo --scope project
- claude plugin install salesforce-development@orgtomo --scope project
- claude plugin install salesforce-code-quality@orgtomo --scope project
- ```
- Then start a session with the Lead and give it a ticket:
- ```bash
- orgtomo start
- ```
- ```
- /orgtomo:ping
- /orgtomo:ticket ABC-123
- ```
- Full walkthrough: [docs/SETUP.md](docs/SETUP.md).
- ## What you type
- | You type | What happens |
- |---|---|
- | `/orgtomo:ticket ABC-123` | The team starts or continues the ticket |
- | `/orgtomo:approve` / `/orgtomo:reject <reason>` | Your decision at a gate |
- | `/orgtomo:answer Q1 <text>` | Answer a blocking question |
- | `/orgtomo:deployed uat` | You finished a deploy; parity check and retest follow |
- | `/orgtomo:feedback bad planner <what you saw>` | Recorded as evidence for lessons |
- | `/orgtomo:rule builder <rule>` | A rule in your words; applies at once |
- | `/orgtomo:models planner opus xhigh` | Model and effort per agent |
- | `/orgtomo:restore ABC-123` | Put the backed-up versions back into the build org |
- | `/orgtomo:status`, `/orgtomo:lessons`, `/orgtomo:brain`, `/orgtomo:doctor` | Look around |
- All commands: [docs/COMMANDS.md](docs/COMMANDS.md).
- ## Configuration
- Everything about your setup lives in `orgtomo.config/` in your workspace. Agents cannot write there.
- | File | Holds |
- |---|---|
- | `orgs.yaml` | Your orgs, their CLI aliases and roles (build, test, reference, production) |
- | `pipeline.yaml` | Build org, deploy targets and tools, loop limit, autonomy, lite path |
- | `safety.yaml` | E-mail mode, allowed test addresses and numbers, masking, browser rules |
- | `agents.yaml` | Model, effort and on/off per agent |
- | `features.yaml` | Lite path, self-learning, meetings, people profiles, Flow generator, telemetry |
- | `conventions.yaml` | Your naming, comment and format rules, and forbidden patterns |
- | `capabilities.yaml` | Which tool provides which capability; add your own MCP servers here |
- Change them in the files, in the control panel (`orgtomo ui`), or with the slash commands. Details: [docs/CONFIGURATION.md](docs/CONFIGURATION.md).
- ## The vault
- Knowledge is kept as Markdown in `vault/`, which opens as an Obsidian vault:
- - `orgs/<org>/`: the org map. One generated note per component with what it uses and what uses it, a hub page per object, a "what runs on this object" note in save order, and a canvas map.
- - `orgs/<org>/business/`: how the business uses the org (optional). Areas and their objects, one process note per area with a fact behind every statement, and the questions only a person can answer.
- - `tickets/<KEY>/`: every stage output, gate report, evidence and deploy pack.
- - `learning/`: observations and lessons.
- - `meetings/`, `people/`: optional features.
- ## Built on
- - [forcedotcom/sf-skills](https://github.com/forcedotcom/sf-skills) (Apache-2.0) for Salesforce skills, the Apex and SOQL language server and Code Analyzer. Orgtomo installs it as separate plugins pinned to one commit (`sf-skills.lock`); nothing from it is copied into this repository.
- - The Salesforce CLI for all org access.
- - [Playwright MCP](https://github.com/microsoft/playwright-mcp) for browser steps (optional).
- ## Status
- Version 0.2.0. Every released version keeps its own branch (`release/v0.1.0`, `release/v0.2.0`), so any of them can be installed by name. The engine, guards and gates are covered by an offline test suite that includes a complete ticket and adversarial cases. What has and has not yet been exercised against live orgs and a live Claude Code session is listed plainly in [docs/VERIFICATION.md](docs/VERIFICATION.md). Read it before relying on Orgtomo for real work.
- ## Documentation
- - [Setup](docs/SETUP.md)
- - [Architecture](docs/ARCHITECTURE.md)
- - [Safety model and its limits](docs/SAFETY.md)
- - [Agents](docs/AGENTS.md)
- - [Knowledge: org map, learning, meetings, people](docs/KNOWLEDGE.md)
- - [Configuration](docs/CONFIGURATION.md)
- - [Commands](docs/COMMANDS.md)
- - [Verification status](docs/VERIFICATION.md)
- ## Contributing
- See [CONTRIBUTING.md](CONTRIBUTING.md). Pull requests target `develop`.
- ## Security
- See [SECURITY.md](SECURITY.md) for how to report a vulnerability, and [docs/SAFETY.md](docs/SAFETY.md) for what the guards do and do not protect against.
- ## License
- MIT. Copyright (c) 2026 Khadda AI Labs. See [LICENSE](LICENSE) and [NOTICE](NOTICE).
Ratings and comments
Sign in to rate this skill.
Sign inFree to use. If it’s helped, pass it on.