What belongs here
This subcategory is for sharing reusable Agent configurations that higher education colleagues can import and adapt for local use. Agents in Alkimi AI export and import as JSON configuration files, so members share a design by posting a description alongside a sanitized JSON export attachment.
Before you submit
Export your Agent configuration JSON and open it in a text editor. Read the file content and confirm that every one of these is gone before you attach it:
- Proprietary policy text that cannot be shared publicly.
- Internal URLs and local server addresses.
- Local contact names, office names, and other identifying details.
- Credentials, including API keys, bearer tokens, passwords, and session secrets.
What an export carries
The JSON carries prompt instructions, model settings, and Safeguard selections. It does not carry linked Knowledge collections.
Adopting institutions build their own collections independently, so describe the source materials and file types your Agent expects. That description is what lets an adopter assemble the equivalent locally.
The required submission format
Opening a topic here pre-fills your composer with the template below, so in most cases you can just fill it in. It is reproduced here for reference:
## Overview
### What it does
[Briefly describe the primary purpose and capabilities of this Agent.]
### Who it is for
[Identify target roles, such as advisors, department chairs, or communications staff.]
### When to use it
[Explain appropriate use cases and scenarios where human staff review is required.]
## Setup
- Source: [Template Catalog, AI Agent Builder, or Blank Agent]
- Recommended model capability: [Standard or Reasoning; specify if reasoning must be disabled]
- Knowledge collections needed: [Describe required source materials and file types]
- Suggested Safeguards: [List active Safeguards, such as Authority Boundary]
- Settings to review after import: [Note settings to check, including citations and search toggles]
### Template or workflow
[Paste core system prompt instructions or workflow guidelines in a code block.]
## In practice
### Example
User request: [Include a representative fictional query]
Expected response: [Provide expected output, including citations or referral placeholders]
### Limitations and safeguards
[Describe operational boundaries, prompt limitations, and required human review steps.]
## Sanitization confirmation
- [ ] I reviewed the exported JSON file in a text editor prior to posting.
- [ ] I removed proprietary institutional policy text that cannot be shared publicly.
- [ ] I removed internal URLs, local server addresses, and non-public links.
- [ ] I removed office names, staff contact details, and local identifying information.
- [ ] I removed student records, employee data, and real case scenarios.
- [ ] I removed credentials, API keys, bearer tokens, passwords, and session secrets.
I confirm the attached export file is safe to share publicly.
## Share your variation
[Invite replies. Name the part you are least sure about, or the adaptation you would find most useful to see.]
What happens after you post
Community moderators read new template submissions, and one arriving without the sanitization checklist completed gets a reply asking for it before other members adopt it. To see what a complete submission looks like, read Example: a campus communications draft assistant.
Where to go instead
- A step-by-step human process with no exported JSON file is still worth posting here. Describe it as a workflow rather than a template, and fill in only the sections of the format that apply.
- General discussion of prompt structure or collection indexing belongs in the parent Discussion category instead, tagged with the product area.
Before you post
Before sharing prompts, configurations, or screenshots, review Community sharing and responsible AI use. Never post student records, personally identifiable information, or credentials.