Anthropic Scheduled Daily Briefing
Suggested workflow: Read Slack and GitHub updates on weekdays and explicitly report if sources are unreadable. Task: Suggested implementation: Configure a scheduled automation agent to run every weekday morning. The agent uses read-only access to fetch updates from Slack and GitHub, relying on per-source bookmarks to track progress rather than fixed time windows. It explicitly reports failed sources as unreadable instead of returning empty results. Human involvement includes defining the initial schedule, managing credential vaults, and reviewing the generated briefs for relevance.; tools: Tool described in the source. Human role: Not stated in the source; limitations: Prerequisites include valid read-only credentials for Slack and GitHub stored in a secure vault outside the agent. The workflow requires a scheduling mechanism (e.g., cron) and a budget cap to prevent runaway costs. Source reliability is secondary, as the text is a public summary of internal practices; specific implementation details may vary. Human approval is necessary for initial setup and ongoing monitoring of source connectivity..
Tools
Tool described in the source
Published on Agent Work Atlas:
Overview
Suggested workflow: Read Slack and GitHub updates on weekdays and explicitly report if sources are unreadable.
Workflow
Suggested implementation: Configure a scheduled automation agent to run every weekday morning. The agent uses read-only access to fetch updates from Slack and GitHub, relying on per-source bookmarks to track progress rather than fixed time windows. It explicitly reports failed sources as unreadable instead of returning empty results. Human involvement includes defining the initial schedule, managing credential vaults, and reviewing the generated briefs for relevance.
Reported result
The source describes an internal Anthropic reference implementation but provides no verified performance metrics, deployment scope, or user satisfaction data for this specific workflow. Actual outcomes such as time saved or error rates are unknown. Proposed acceptance check: Verify that the agent successfully retrieves data from at least 95% of configured sources over a one-week pilot and correctly flags any inaccessible sources as 'unreadable' in the output.
Human role
Not stated in the source
Limitations
Prerequisites include valid read-only credentials for Slack and GitHub stored in a secure vault outside the agent. The workflow requires a scheduling mechanism (e.g., cron) and a budget cap to prevent runaway costs. Source reliability is secondary, as the text is a public summary of internal practices; specific implementation details may vary. Human approval is necessary for initial setup and ongoing monitoring of source connectivity.
Try it
Act as an automation engineer. I need to set up a daily briefing agent for my team. Context: We use Slack for communication and GitHub for code. I want a summary of key updates every weekday morning. Inputs: - Slack Workspace URL: [replaceable_slack_url] - GitHub Organization: [replaceable_github_org] - Target Channel for Output: [replaceable_output_channel] - Budget Cap: $5 per run Steps: 1. Create a configuration file for a scheduled agent that runs at 8:00 AM on weekdays. 2. Define read-only access scopes for Slack and GitHub, ensuring credentials are stored in a vault, not in the prompt. 3. Implement logic to use per-source bookmarks to track the last read position, avoiding duplicate reports. 4. Configure error handling: if a source is inaccessible, explicitly state 'Source Unreadable' in the report rather than omitting it. 5. Set a hard budget cap of $5 per execution. Deliverables: - A YAML or JSON configuration file for the agent. - A sample output format showing how updates and errors are displayed. Quality Check: - Run a dry run with a test dataset. - Verify that the output correctly identifies unreadable sources. - Confirm the budget cap is enforced in the configuration. Human Confirmation: - Do not deploy to production. - Present the configuration and sample output to me for review before scheduling the actual task.
Source reports are not independently verified. Publication date is the catalog publication date, not the original article date.