LeslieWylie/agent-loop-workflow
agent-loop-workflow: Universal multi-agent collaboration workflow skeleton skill plug-in — Loop Guard/Handoff/Review→Clos
agent-loop-workflow is 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. Its modules are: a four-role topology (DRI/doer, reviewer, lead for decision/routing, ops for automation) with clear responsibility boundaries; a Loop Guard with six invariants (turn limit, progress-stall detection, repeated-error fuse, timeout warning, an explicit exit gate, and a single authoritative writer); a six-field standard handoff template (owner, goal, input, permitted writes, acceptance criteria, failure evidence); risk-based routing across fast/standard/high lanes; a fixed delivery sequence (verify -> commit -> push -> Draft MR/PR -> ready signal -> in_review); a review-to-close protocol; anti-loop protection (idempotent issue creation, state-change filtering, human-source-only new-task trigger); and red lines (secret isolation, token permission separation, destructive-operation confirmation).
Install
dsh plugin --profile web add "github:LeslieWylie/agent-loop-workflow"It is installed as a GitHub source skill-plugin with the line above; after restart, load the skill in any session with load agent-loop-workflow. The README notes it has zero dependencies -- no external services, only the DSH skill-loading mechanism. No npm registry package is claimed and none was checked.
Compatibility
DeepSeek Harness. Project-agnostic and forge-agnostic: the skill defines collaboration flow rules only and maps its "issue"/"MR" onto whatever tracker and forge you already use. Zero dependencies (no external services; only the DSH skill-loading mechanism). Project-specific knowledge is deliberately left to separate -conventions / -engineering / *-review-rules skills.
Details
- Repo: LeslieWylie/agent-loop-workflow
- Category: Agent Capabilities
- Stars: 1
- Version: GitHub source install (no npm registry package claimed); MIT
- Last push: 2026-08-17
- First seen: 2026-08-14
Recent updates
The README documents the seven capability modules as a table and three design principles: project-agnostic (only collaboration flow rules; project knowledge goes in separate skills), zero dependencies (external services not required, only the DSH skill-loading mechanism), and forge-agnostic (the skill names no tracker or forge and maps onto whatever you use).
FAQ
- How do I install agent-loop-workflow?
- Run dsh plugin --profile web add "github:LeslieWylie/agent-loop-workflow", restart, then load it in any session with load agent-loop-workflow.
- Does it lock me to a particular tracker or forge?
- No -- the README states the skill is forge-agnostic: it describes the collaboration flow only and maps its "issue" and "MR" onto whatever tracker and forge you already use.
- What does it need to run?
- Nothing extra: per the README it has zero dependencies and requires no external services, only the DSH skill-loading mechanism.
Alternatives
february2015/dsh-taskswarm · akqwpeter-prog/dsh-agent-conductor · aaravarr/dsh-subagent-max