Let an agent extend its own code inside a Docker sandbox
From the linked report, miguel, assembled on Claude and the Agno framework, runs inside Docker with the project mounted read-only and only its own code directory writable. A host-side runner snapshots git, gives the agent its source plus the next capability, then validates changes with AST, JSON schema and import checks. It finished all 10 seed capabilities, wrote its own checklist, and reached 21+ capabilities.
- Agent or tool
- Claude
- Topic
- DevOps & Infra
- Reported on
- Hacker News
- Added
- 2026-10-15
Why it matters
Keeping servers and pipelines healthy is constant, detail-heavy work. Agents that can read logs and run routine fixes are a big help, as long as a person keeps an eye on what changes. It is one of 130 sightings filed under DevOps & Infra on Agent Sightings (number 124 by interest), and one of 213 involving Claude.
How you could try it
Here is a reusable prompt, written by us in our own words, that captures the task. Paste it into Claude or any similar agent and adjust the details to your situation.
Infra task — let an agent extend its own code inside a Docker sandbox. Inspect current config, propose a diff, and wait for approval before applying or restarting anything.
Tip: Try it first on a staging system or with read-only credentials, and require approval for anything that changes production.
Source
Read the original first-hand account on Hacker News: https://news.ycombinator.com/item?id=47329087. Our summary is a paraphrase; details are in the source.
Built an AI agent? Feature your agent on Agent Sightings, free submission or a featured spot.
Browse more like this
Related sightings
- Model Experimentation Made Easier
- Qualified Health and the University of Texas System use Claude to identify patients who need life-saving care
- Harvey transforms legal work with Claude
- Panorama unlocks student success with Claude in Amazon Bedrock
- Cut a hospital bill using Claude's Medicare price analysis
- Run your side of a home insurance claim