Skip to content

About

Build reviewable prompt context from Salesforce records without sending customer data to an AI model.

Resources

Code of conduct

Contributing

Security policy

Stars

0 stars

Watchers

0 watching

Forks

Latest commit

 

History

21 Commits

Folders and files

NameName
Last commit message
Last commit date
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Prompt Context Builder

Quality Salesforce API 67.0

Build reviewable prompt context from Salesforce records without sending customer data to an AI model.

Warning

Active development: This project is pre-release and may change in breaking ways without notice. Package contents, APIs, configuration, data models, and upgrade behavior are not yet stable. Use it only in development or sandbox environments, expect that updates may require reconfiguration or reinstallation, and do not rely on it for production workloads yet.

Prompt Context Builder lets an administrator create a reusable prompt template, choose the current record fields and limited related data it may read, preview the result, and publish a fixed version. Users can then generate, edit, and copy the completed plain-text prompt from a record page. Record pages and Salesforce automation use the same access checks.

It does not call an AI model, read Salesforce File bodies, or update customer records. Flow and Apex can explicitly save prepared prompt text plus file metadata to a private Prompt Input Log; record-page generation remains an edit-and-copy workflow and does not save the prompt.

Get started

Try the prepared scratch-org demo · Deploy source to a sandbox · Create your first template · Browse all documentation

The current public evaluation path is source deployment. A public package installer is not advertised until the release owner approves an exact promoted version. Package aliases and build records are release evidence, not a promise that a version is publicly supported.

See it in action

Published template open in Prompt Builder

Deployed scratch-org screenshot, September 26, 2026; see screenshot evidence.

The prepared demo creates the same Account and Contacts each time, along with 12 numbered templates. Nine templates are published and runnable; two remain Draft to teach validation behavior, and one is Inactive to demonstrate lifecycle enforcement.

What to explore What the demo proves
Merge syntax Current record fields, indexed records, counts, loops, conditions, fallback text, and formatting
Related and reference data sets Current-record matches and limited shared lists
Filters, sorting, and limits Field-aware rules, named groups, repeatable ordering, and separate record limits
Calculated values Counts, sums, averages, minimums, and maximums
Safe failure Invalid Draft, Inactive template, empty data, and inaccessible-data behavior

See the current and historical UI evidence and the verification catalog for every prepared scenario.

How it works

  1. An administrator creates a Draft template for one Salesforce object.
  2. The administrator chooses current-record fields and up to ten related data sets or calculated values.
  3. Validation checks the prompt text, Salesforce fields, required access, record order, and limits.
  4. Preview runs against a real record with the administrator's current Salesforce access.
  5. Publish creates a fixed version for record pages and automation.
  6. Generation reads only data visible to the running user and either returns the complete prompt or returns no prompt when any required part fails.

Administrators choose objects, fields, filters, sorting, and limits through Prompt Builder. The builder creates the Salesforce searches from those choices.

Account: {!record.Name fallback="Unknown account"}
Open contacts: {!related.openContacts.count}

{!#each related.openContacts as="contact"}
- {!contact.Name}: {!contact.Email fallback="No email"}
{!else}
No open contacts were found.
{!/each}

Try the prepared demo

You need Git, Node.js 22, Salesforce CLI with the Code Analyzer plugin (sf plugins install code-analyzer), and access to a Dev Hub that can create scratch orgs. Creating a scratch org consumes Dev Hub capacity. Run these commands only with the Dev Hub owner's approval, replacing TARGET_DEV_HUB with an authenticated alias:

git clone https://github.com/gkolan/prompt-context-builder.git
cd prompt-context-builder
npm ci
npm run ci:gates

sf org create scratch \
  --definition-file config/promptcontextbuilder-scratch-def.json \
  --target-dev-hub TARGET_DEV_HUB \
  --alias pcb-demo \
  --duration-days 7

sf project deploy start --source-dir force-app --target-org pcb-demo --wait 30
sf org assign permset --name Prompt_Context_Configure --target-org pcb-demo
npm run demo:setup-source -- --target-org pcb-demo
npm run demo:verify-source -- --target-org pcb-demo
sf org open --target-org pcb-demo --path lightning/n/Prompt_Builder

The verifier succeeds only when all 12 teaching templates have the expected lifecycle state and all nine published examples generate nonblank prompts. Continue with the complete demo guide for expected output, UI steps, ordinary-user verification, troubleshooting, and cleanup responsibility.

If you have a sandbox but no Dev Hub, follow deploy source to a sandbox.

What is included

Area Included behavior
Template lifecycle Draft, validate, preview, publish, compare versions, restore to Draft, and deactivate
Related data Several related or reference data sets and calculated values
Record choices Salesforce fields, filters, named groups, repeatable sorting, and record limits
Ways to use it Lightning record pages, Flow, Apex, Batch, Queueable, and Scheduled Apex
Data access Salesforce sharing and field access, with no partial prompt after a failure
Version history Stable automation keys, fixed published versions, signatures, and scheduled checks
Product checks Salesforce and component tests, source review, sample data, and 12 teaching templates

For exact syntax, permissions, limits, reason codes, and API contracts, use the reference documentation.

Security model

The package Permission Sets authorize Prompt Context Builder features; they do not grant access to business objects, fields, or records. Salesforce sharing, restriction rules, object access, and field access continue to control every record the template reads. A missing permission is an error, not a blank value, and a failed request returns no partial prompt.

Start with security and data access and Permission Sets before production evaluation.

Documentation

Goal Guide
Decide whether the product fits Start here
Try a prepared source deployment Scratch-org demo
Deploy to an existing sandbox Install
Create and publish a template Administrator walkthrough
Configure the record-page component Lightning record page
Call from Flow or Apex Flow and Apex
Troubleshoot a safe reason code Diagnostics
Review exact contracts Reference
Operate and monitor the capability Production operations
Upgrade source or a package Upgrade
Remove an evaluation or installation Remove safely

Contributing and support

  • Read Contributing before opening a pull request.
  • Use Support for safe issue-intake guidance.
  • Report vulnerabilities privately under the Security policy.
  • Run npm run ci:gates before handing off a change.

License

No open-source license has been selected. Before describing this project as open source, the release owner must add the approved license and copyright holder. Until then, publication of the source does not grant reuse rights.

About

Build reviewable prompt context from Salesforce records without sending customer data to an AI model.

Resources

Code of conduct

Contributing

Security policy

Stars

0 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages