Engineer shares minimalist Claude-based workflow for building AI agents
Start here
An engineer argues that building AI agents has come full circle: after years of frameworks, orchestration engines, multi-agent architectures and vector databases, the work now reduces to a folder of Markdown instructions driven by Claude. The workflow starts with `mkdir`, an AGENTS.md file and a conversation with Claude, adding only the skills, context and smallest set of tools the job needs, then invoking it headless with the `-p` flag. Headless means there is no interactive chat, and the author says tool permissions are configured in advance for unattended work; Claude can also return structured JSON for another program to use. Session transcripts under ~/.claude/projects/ are used to diagnose failures, while the author says evals remain the hardest part.
Generated from reports · updated 16 h ago
Report timeline
Follow the reports to see the event from different sides.
- Hacker News · AIWhat Picasso Taught Me About Building AI Agents
An engineer argues that building AI agents has come full circle: after years of frameworks, orchestration engines, multi-agent architectures and vector databases, the work now reduces to a folder of Markdown instructions driven by Claude. The workflow starts with `mkdir`, an AGENTS.md file and a conversation with Claude, adding only the skills, context and smallest set of tools the job needs, then invoking it headless with the `-p` flag. Session transcripts under ~/.claude/projects/ are used to diagnose failures, while evals remain the hardest part.
Heat for this event
- Current, comparable range
- 6
- Peak, comparable range
- 10November 16 01:00
- Change in 24 hours
- –
The trend compares the same participants over hours observed continuously, so its range can be narrower than the current heat count. Hover or tap the chart for hourly heat; the arrow keys move through it.