Example Context Packs

Some examples of potentially useful packs:

Daily social media pre-filler

  • Description: Grabs latest dailies from the content calendar Google Sheet and generates a preliminary post about the content to be reviewed. Allows pushing directly to platforms or via Buffer, depending on prompts. Uses the canonical post template as the resource context to generate style and tone of the posts.
  • Associated capabilities:
    • Tools: schedule-post-on-buffer, access-airtable-content-planning-document, post-on-[platform] (platforms not supported by Buffer).
    • Resources: an-example-social-media-post-[platform] (one for each wanted platform)
    • Prompts: create-social-media-posts-prompt-template
  • Usage example: Use the configured prompt template to retrieve content plan from Airtable. The template should instruct the LLM to use the configured example social media post resource as context and produce a post based on that, but using the wanted content from Airtable. Finally produced content should be scheduled onto various platforms using the configured tools in the pack.

Daily security nudge

  • Description: Grabs information from SecureFrame regarding audit status and failing SOC2 tests. Uses team information from Backstage to send a Slack message to the offending teams.
  • Associated Capabilities:
    • Tools: retrieve-secureframe-audit-test-status, send-message-to-slack, query-backstage-component-owner
    • Resources: template-nudge-message
    • Prompts: N/A
  • Usage example: An AI Agent (or a human manually using an LLM client) is configured to run daily in the mornings to retrieve SecureFrame test statuses and associate the components behind these failing tests with their owners, using Backstage service catalog's relationships. Once identified, the agent uses the provided template message as context and produces personalized messages to each culprit. Finally the Slack messaging tool is used by the LLM to send a message to teams or individuals to rectify the issues.

Lead validation pack

  • Description: A pack retrieving user information based on received email and finding information about them. Uses Perplexity and a locally hosted, connected MCP server to find more information about the messager and the company they work for. Storing the retrieved information to SalesForce
  • Associated Capabilities:
    • Tools: linkedin-scraper (Local MCP tool), search-user-with-perplexity, store-record-to-salesforce,
    • Resources: salesforce-record-structure
    • Prompts: N/A
  • Usage example: Sales person receives a notification about an interest in the product. They trigger an LLM to find more information about the user. The LLM uses locally connected linkedin-scraper to retrieve initial information about the person. Subsequently, the LLM uses Perplexity API endpoint wrapped into a tool to find more specific information outside Linkedin about the user and their company. Finally, they record contact information and short bio of the person to SalesForce.