DeepSeek Harness Plugin

LeslieWylie/agent-loop-workflow

Stars ★ 1 Category Workflow & Automation Added 2026-08-17

A project-agnostic multi-agent collaboration protocol for the DeepSeek Harness: loop guards that stop runaway agent loops, a six-field handoff format, risk-tiered review routing, and a fixed verify→commit→push→review→close sequence. One skill instead of the same rules copy-pasted into every agent's instructions.

Install

# from GitHub (first run asks for allowBuilds approval — follow the hint, retry)

dsh plugin --profile web add github:LeslieWylie/agent-loop-workflow

Any plugin you install runs third-party code with your own permissions — it can read your files, use your credentials, and reach the network, and tool approvals don’t sandbox it. GitHub-sourced plugins also run build scripts at install time — pnpm blocks those until you allow them, so an install can stop with ERR_PNPM_GIT_DEP_PREPARE_NOT_ALLOWED or ERR_PNPM_IGNORED_BUILDS; dsh prints the exact key to add under allowBuilds in your profile’s pnpm-workspace.yaml, and the install works on the next run. Allowing a build is a trust decision: only install sources you trust, and pin a commit (github:owner/repo#sha).

README

简体中文

What is this

A project-agnostic multi-agent collaboration workflow skeleton. Any agent squad composed of DRI + reviewer + lead + ops can load it to get consistent workflow rules, without duplicating them across every agent's instructions.

Core Capabilities

Module Content
Role Topology DRI (doer), reviewer, lead (decision/routing), ops (automation) — four roles with clear responsibility boundaries
Loop Guard (6 invariants) Turn limit, progress stall detection, repeated error fuse, timeout warning, explicit exit gate, single authoritative writer
Standard Handoff Format 6-field handoff template: owner, goal, input, permitted writes, acceptance criteria, failure evidence
Risk-based Routing fast / standard / high lanes, routing to different reviewers based on change scope
Delivery Sequence verify → commit → push → Draft MR/PR → ready signal → in_review (fixed order)
Review→Close Protocol Automatic in_review dispatch, reviewer metadata, reject/retry, escalation
Anti-loop Protection Idempotent issue creation, state-change filtering, human-source-only new task trigger
Red Lines Secret isolation, token permission separation, destructive operation confirmation

Install

dsh plugin --profile web add "github:LeslieWylie/agent-loop-workflow"

After restart, load the skill in any session:

load agent-loop-workflow

Design Principles

  • Project-agnostic: Only defines collaboration flow rules. Project-specific knowledge goes in *-conventions / *-engineering / *-review-rules skills
  • Zero dependencies: No external services required, only the DSH skill loading mechanism
  • Forge-agnostic: the skill describes the collaboration flow only. Map its "issue" and "MR" onto whatever tracker and forge you already use — it names none.

License

MIT

Content from the project README on GitHub ↗

Links

More in this category

View the whole category →

Community comments

Comments are public GitHub Discussions. Loading them connects to GitHub and Giscus; a GitHub account is required to post.