Saving Specs

Capture successful workflows on the fly with the cospec_save_spec tool. Turn dynamic, intuitive agent work into reusable specs automatically.

💾What is cospec_save_spec?
A tool for programmatically creating specs from successful workflows

The cospec_save_spec tool allows your agent to save entire workflows that were performed successfully on the fly—even if they weren't previously spec'd. Most work is dynamic and intuitive, and this tool captures those moments of success.

How it works

When your agent completes a task using various tools across different platforms, it can use cospec_save_spec to save the entire workflow—including all the tool calls it made to each platform—as a new spec.

Tool name: cospec_save_spec
Parameters:
  • title (required): The title for your new spec
  • content (required): Spec content with text and slash commands
  • description (optional): Optional description for the spec
⚡When to use it
Capture successful workflows in real-time for future reuse

Most work happens dynamically—you don't always know the exact workflow you'll need until you're in the middle of solving a problem. When your agent successfully completes a complex task, save it for future use.

After successful multi-platform workflows

Your agent just coordinated across GitHub, Slack, and Linear to resolve an issue. Save this workflow so it can be reused for similar issues.

When discovering new patterns

Your agent found an effective way to handle customer feedback by pulling from multiple sources and creating a summary in Notion. Capture this for next time.

Building reusable automation

Your agent completed a complex data migration successfully. Save the workflow so you can repeat it with confidence for the remaining migrations.

📝Example usage
See how agents save workflows programmatically

Scenario: Bug fix workflow

Task: Your agent was asked to "fix the login bug"

What happened: The agent performed these actions:

  • Used /github_get_issue to fetch issue details
  • Used /github_search_code to find related code
  • Made the fix and created a PR
  • Used /slack_post_message to notify the team

Agent saves it:

cospec_save_spec(
  title: "Bug fix workflow",
  description: "Standard workflow for fixing bugs from GitHub",
  content: "
    Fix a bug from a GitHub issue:
    1. Get issue details with /github_get_issue
    2. Search for related code with /github_search_code
    3. Create a fix and PR
    4. Notify team on /slack_post_message
  "
)

Scenario: Weekly report generation

Task: Your agent created a weekly status report

What happened: The agent:

  • Used /linear_list_issues to get completed tasks
  • Used /github_list_pull_requests to see merged PRs
  • Used /google_docs_create to generate a report
  • Used /slack_post_message to share with the team

Agent saves it:

cospec_save_spec(
  title: "Weekly status report",
  description: "Generate and share weekly progress",
  content: "
    Create weekly status report:
    1. Get completed tasks from /linear_list_issues
    2. Check merged PRs with /github_list_pull_requests
    3. Generate report with /google_docs_create
    4. Share on Slack with /slack_post_message
  "
)
🔄Building a library of workflows
Turn one-time successes into repeatable processes

Every time your agent successfully completes a task, it has the opportunity to save that workflow. Over time, you build a library of proven patterns that can be reused, refined, and shared.

🎯 Capture success

When something works well, save it immediately so you can repeat the success

🔁 Reuse patterns

Build a collection of workflows that your agent can reference and use again

📚 Grow your toolkit

Each saved spec adds to your organization's knowledge base

🚀 Scale faster

New team members benefit from proven workflows saved by others