Description
Run several coding agents at once and you end up dispatching, nudging, verifying and merging by hand, never sure who skipped the evidence. dori-superagent turns those steps into commands and loops: one lead session talks to you, each task runs as its own lane from a written brief, and finishing requires evidence; a PR that doesn't qualify can't be merged.
It's about 1,600 lines of TypeScript running on herdr and omo, under the MIT licence, plus a Dori skill for chat tools such as Slack, Discord and Telegram.
Lead plus lanes: One lead session talks to you; each task gets its own session in a herdr tab, started from a written brief.
Duty loop: When a lane goes quiet the loop nudges it and escalates to the lead if needed.
Evidence to finish: The READY message must carry evidence, otherwise the task isn't done.
Merge gate: Whether a PR can merge is decided by code, not manual checking.
Dispatch from chat: The Dori skill lets you ask in chat, with one status message per task edited in place.
It's about 1,600 lines of TypeScript running on herdr and omo, under the MIT licence, plus a Dori skill for chat tools such as Slack, Discord and Telegram.
Features
Lead plus lanes: One lead session talks to you; each task gets its own session in a herdr tab, started from a written brief.
Duty loop: When a lane goes quiet the loop nudges it and escalates to the lead if needed.
Evidence to finish: The READY message must carry evidence, otherwise the task isn't done.
Merge gate: Whether a PR can merge is decided by code, not manual checking.
Dispatch from chat: The Dori skill lets you ask in chat, with one status message per task edited in place.
