Prompt Library

Prompt Library#

You drive Squirro CLI through an AI coding assistant, so most work starts with a sentence rather than a command. This page collects prompts for common delivery tasks and groups them by the part of the platform they configure.

Use them as starting points. Replace the placeholders in angle brackets with your own project directory names, agent names, and values, and read the preview the assistant shows you before confirming any apply. For the commands behind these prompts, see the Main Commands page.

Note

Start your assistant from inside the configuration directory, so that it picks up the CLAUDE.md file and the skills that squirro init wrote. For more information, see the Configuration Directory page.

Configuration and Environments#

Setting up and moving configuration between instances.

Initialize a new Squirro configuration directory in the current folder and connect it to our staging environment.

Create a configuration directory called <name> and register the Squirro project <project-id> under the <profile> profile.

Pull the full configuration from our reference project so I can use it as a baseline for the new engagement.

Create separate profiles for <name>-dev, <name>-staging, and <name>-prod, each with its own cluster URL.

Show me all profiles and their validation status.

Show the details of the <profile> profile, including the projects it maps.

Test my connection to the instance and confirm I have the right permissions.

Check my connection to both GenAI Platform and Graphite for the <profile> profile.

These prompts promote a change from one environment to the next.

Validate all configuration files before I apply them to production.

Dry-run the apply so I can review every change without making anything live.

Apply the configuration from the <profile> profile to staging, matching records by name, and dry-run it first.

Apply the <name> baseline configuration to the new project on the development instance.

Platform and Access#

Project settings, theming, members, and server-wide groups.

Apply the updated theme configuration to staging, and dry-run it first.

Apply the members configuration to the new project to set up team access.

Apply the updated groups configuration to the server, previewing the changes first.

Validate the platform configuration files before applying them to production.

Project-level and server-level settings live in config-project.json and config-server.json, which are free-form key-value stores covering query templates, chat preferences, and feature flags.

Create a query template that filters all items with source type Microsoft SharePoint.

Add a query template that filters results by the user group identifier from the single sign-on attributes.

Update the query template to exclude items with processing errors from all search results.

Update the chat widget colors to match our brand, with primary color <hex> and icon color <hex>.

Turn off starter questions in the chat widget for the production project.

Set the chat widget to the left side and add three starter questions for the onboarding project.

Set includeSources in the chat preferences so that users always see document citations.

Turn off the notes feature for the production project.

Turn on the copilot chat feature flag for the pilot project.

Data Sources#

Connector configuration and loading data.

Set up the prerequisites for the SharePoint connector. I have already stored the client id and the client secret myself, so do not ask me for them.

Apply the SharePoint connector configuration to the new project.

Apply the updated connectors configuration, including the new Confluence source, to staging, and dry-run it first.

Apply the labels configuration and all source configurations to the project.

Validate the connector configuration files before applying them.

Load <file> into the file-based source <source-id> and wait for the run to finish.

Enrichment Pipelines#

Pipeline workflows and pipelets.

Read the current enrichments/workflows.json for project <folder>, then add a workflow named <name> with these steps: <describe the steps>. Leave the id null so the instance assigns one. Preview with a dry run, then apply.

In projects/<folder>/enrichments/workflows.json, rename the workflow <old name> to <new name>. Leave the id field alone so the change applies as an update. Preview, then apply.

Add an optical character recognition step to the SharePoint workflow.

Reprocess every document through the workflow <workflow-id> for project <folder>.

Apply only the enrichments configuration to the new instance.

Validate the enrichment workflow configuration files before applying them to production.

Pipelets are Python plugins that process items after fetching and before indexing. Squirro CLI bundles a skill for building them, so an assistant can write the pipelet, lay out the directory it needs, and take it through to deployment. The upload itself runs through the Squirro Toolbox, using pipelet upload or squirro_asset, so install the Toolbox as well if you work on pipelets. For more information, see the Squirro Toolbox page.

Create a pipelet that tags items with a status keyword based on keywords found in the body.

Create a pipelet that calls an external API to classify documents before indexing.

Create a pipelet that extracts named entities from the document body and stores them as keywords.

Create and deploy a pipelet that filters out items with no body before indexing.

Agents#

Agent configuration, personas, and live testing.

Apply the updated agent configuration to staging, and dry-run it first.

Add a new tool to the <agent> toolkit and apply the change.

Turn off the <agent> agent by setting active to false, then apply to production.

Copy the agents directory from project <source> into project <target> and apply.

Pull the <agent> configuration and make its persona more concise and domain-specific.

Make the <agent> persona more neutral and less technical for end users, then dry-run to preview the change.

Pull all agent configurations for project <folder> and align their personas to a consistent voice.

Note

The persona field accepts at most 3999 non-whitespace characters. Squirro CLI does not check the length, so an over-long persona passes config apply --dry-run cleanly and is refused by the instance once you confirm the apply.

These prompts test agents against a live instance.

List all agents configured in the current project.

List the agents in JSON format so I can pipe them into another tool.

Send a test message to the <agent> agent to check that its persona and grounding work.

Start an interactive chat session with the <agent> agent, with streaming turned on.

Send "<question>" to the <agent> agent and return the response as JSON.

Diagnostics#

Inspecting the API surface and reading instance logs.

List all available API methods for the topic service.

Call get_project for project <project-id> and show the full JSON response.

Make a raw GET request to /api/v0/users/me to check my authentication.

List every configuration type Squirro CLI can manage.

Generate the JSON schema for the agents configuration file so I can add editor validation.

Fetch the last 50 error-level entries from the <log> log.

Get the entries from the <log> log for the last hour, filtered to warning level and above.

Next Steps#

For the commands these prompts run, see the Main Commands page. To suggest a prompt for this page or to report a task Squirro CLI does not yet cover, visit the Squirro Support website and submit a feedback request.


Contents
This page
View as Markdown