YT · YOUTUBE// 25d agoOPENSOURCE RELEASE

do-work brings queue discipline to autonomous coding

do-work is an open-source skill that splits agentic coding into a fast capture phase and a structured execution loop, so teams can log requests immediately and process them later with less context switching. In the demonstrated workflow, each request is triaged by complexity, executed in order, and archived with implementation notes plus per-request git commit traceability.

// ANALYSIS

The strongest idea here is not “more autonomy,” but operational control: do-work makes long-running agent sessions auditable instead of ad hoc.

–The capture/process split supports a real “walk away” workflow without losing the original user intent.
–Queue files (`REQ-*`) and user request folders (`UR-*`) create a durable paper trail from prompt to shipped change.
–Complexity triage (simple/medium/complex) prevents over-planning tiny fixes while forcing rigor for architectural work.
–Atomic, per-request commits make rollback, review, and handoff much cleaner than one giant AI-generated diff.
–The model is tool-agnostic enough to transfer beyond Claude Code into other agentic coding environments.

// TAGS

do-workai-codingagentautomationclidevtoolopen-source

DISCOVERED

25d ago

2026-03-17

PUBLISHED

25d ago

2026-03-17

RELEVANCE

7/ 10

AUTHOR

Matt Maher