ggondim

Autoducks — AI skill for Claude Code

AI community

autoducks — run your own AI agent harness inside GitHub, via /agents verbs.

How to install Autoducks

This entry records only its repository, not the path inside it, so there is no exact command to give. Open ggondim/claude-in-github and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Autoducks does

autoducks — run your own AI agent harness inside GitHub, via /agents verbs. BYO subscription, deterministic orchestration, chatops-native.

Alternatives in AI

Osaurus — Own your AI. The native macOS harness for AI agents -- any model, persistent memory, autonomous execution, cry 5.1k ★
TokenTracker — Local-first AI token usage & cost tracker for 31 coding tools incl 1.4k ★
Loushang — AI-native agent harness for coding workflows by python: multi-model LLM orchestration, stateful sessions, tool 1.3k ★

README

autoducks

autoducks lets you run code agents on your CI/CD platform, triggered by issue comments, with any LLM provider. One command drives a full design → plan → execute pipeline: each agent checks its own Definition of Ready and automatically dispatches the missing prerequisite, so you always get a reviewed design and plan before code is written.

Full documentation lives at ****.

Four agents, one pipeline

Agent	Layer	What it does	Trigger phrases
Architect	Design	Creates or revises the design of a feature/bug in the issue body	`/architect`, `/design`
Engineer	Tactics	Breaks the design into task issues organized in dependency waves	`/engineer`, `/tactics`
Maestro	Orchestration	Owns branches/PRs and dispatches execution waves in parallel	`/execute`, `/run`, `/work`
Developer	Build	Implements one task and merges its PR into the pipeline branch	(dispatched by the Maestro, or `execute` on a Task issue)

The same `execute` comment routes to the right agent by issue state — you never have to know which workflow runs. On a raw issue it cascades through the whole pipeline automatically (Architect → Engineer → Maestro → Developers). You can also chain agents explicitly: `/architect #auto:engineer+execute`.

Utility agents: `/review` (review a finished PR against its design and acceptance criteria — read-only, never merges), `/rework` (distill unresolved review feedback into one follow-up task and revert the PR to draft), `/defer` (capture unresolved review feedback as a follow-up issue so the PR can be merged/closed now), `/triage` (groom the backlog: assign priorities, propose duplicate groupings — scheduled, on issue open, or on demand), `/merge` (deterministically close one issue as a duplicate of another), `/fix` (repair a failed task run), `/revert` (undo a feature, restore the human-authored issue), `/close`