Learning Timeline
Key Insights
Benefits of Tracing over Background Jobs
Unlike standard 'cron jobs' that run in the background without visual logs, Paperclip Routines allow you to 'trace' every task. You can monitor token usage and see exactly what the agent did to help with continuous improvement.
Token-Saving Tips
Always include the instruction 'do nothing if there isn't anything' in your prompt. This prevents the AI agent from generating unnecessary output or hallucinating when no activity is detected, saving you on API costs.
Prompts
Content Strategist Agent Instructions (GitHub to Discord)
Target:
Paperclip Agent Prompt
Every day we put work into paperclip. We're merging pull requests in a community formatted form. We especially want to celebrate our community members who contribute. So, make sure you call out everyone who added something that was merged. We do take the weekends off sometimes, so if there isn't anything, you can just do nothing. Read the GitHub changes for the repo. Eventually we'll post direct to Discord but for now just post in your issue.
Step by Step
Creating an AI Agent Automation Routine
- Open the Paperclip dashboard and either select or create an AI agent (e.g., Content Strategist).
- In the agent configuration section, enter the task description and project context in the instructions area.
- Write specific instructions for the agent to check for code changes (pull requests) in the relevant GitHub repository.
- Set up detection logic: Instruct the agent to focus on code merged into the 'main' or 'master branch'.
- Add instructions to 'celebrate' or mention the names of community contributors involved in the updates.
- Define exception rules (e.g., if there is no activity over the weekend, the agent should do nothing).
- Click the 'Create Routine' button to save your task settings.
- Click the 'Trigger' menu to schedule when the automation should run.
- Choose a frequency (e.g., 'Every day') and set a specific time (e.g., '10 a.m.').
- Click 'Add Trigger' to activate the daily routine.
- Check the logs or 'Run' history to see the details of completed tasks, including token usage and the agent's findings.
