ADLC Harness plugin
ADLC Harness takes a task from intent to PR with one command, and the human approves only four key gates.
- 125 ★ on the repository
- Last commit
Extended access
Why this level
- The plugin installs hooks, writes documents and commits into the repository and keeps memory in the home folder.
- Release (push and PR) needs approval, while other stages run without questions.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add plugins/adlc-harnessDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- The plugin installs hooks, writes documents and commits into the repository and keeps memory in the home folder.
- Release (push and PR) needs approval, while other stages run without questions.
This is third-party code. Review the repository files before installing.
What it does
A plugin for Claude Code, Codex and GitHub Copilot CLI with 29 skills, three subagents and two hooks, with no runtime dependencies. The `/loopeng <request>` command runs the loop intent, design, plan, build, verify, review, ship, learn. The human approves intent, design, plan and release (push and PR). Each stage leaves a document and a git-recorded approval. Memory lives in one folder shared by all three hosts: short-term in `handoff.md`, mid-term in indexed notes, long-term in `AGENTS.md`. Review runs in parallel by lens with a verifier vote.
Who it is for. Tech leads and developers running work across several CLI agents who want shared discipline: documents, approvals, verification and memory.
Good fit when
- A feature goes from idea to PR and each stage must leave a document.
- Work moves between Claude Code and Codex without losing its place.
- Code review needs parallel passes and verification of findings.
- Agents need shared memory of project decisions and mistakes.
Not a fit when
- The task is a couple of lines and the full document loop is overkill.
- The environment lacks Node.js 22 or the right to install plugins.
- The team does not want the agent committing documents to the repository.
Example request
/loopeng add CSV report export to the service: start with intent and stop at design approval.Limitations
Node.js 22 or newer is required. Codex hooks need a one-time trust in `/hooks`. The documentation is in Korean, while skill names are English. The loop creates documents and commits under `docs/work/`, so the repository gains many service files. Three consecutive failures from one cause stop the run. The repository states no license. Installing RTK and Impeccable from the README is optional and adds third-party hooks.
How to disable. Run `claude plugin uninstall adlc-harness@adlc-harness` or the Codex and Copilot CLI equivalent. Delete `docs/work/` folders and memory under `~/.claude/projects/` by hand.
FAQ
When does the loop stop for a human?
At intent, design, plan and release. Build, verify and review run without questions.
How does memory work?
Short-term sits in `handoff.md`, mid-term in notes indexed by `MEMORY.md`, long-term in the project `AGENTS.md`. Repeated corrections move up.
Which hosts are supported?
Claude Code, Codex and GitHub Copilot CLI, each with its own `marketplace add` and install commands.
What about version 1.x?
The old version copied files into `~/.claude`. The README advises backing up and removing the copies before installing the plugin.
How do I resume interrupted work?
Call `/loopeng`: state is restored from documents, approval lines and plan checkmarks.
Discussion
No questions or reviews yet.
Related
PluginsHooks and skills make Claude Code plan before coding, test before shipping and escalate when it is not confident.
A DiDi plugin for Codex and Claude Code that runs a task from proposal to archive and keeps verified project knowledge in the repo
A Claude Code plugin with /readiness and /setup skills that scores a repo for agent work and installs rules, hooks and checks.
The AI-SDLC plugin for Claude Code carries a task from spec to PR: a readiness gate, a developer agent and several reviewers.
Harness drives a task through plan, work and review: spec and Plans.md live in the repo and a separate reviewer checks the result.
ADLC is a set of dependency-free CLIs and agent plugins: gates at each development phase against typical model failure modes.
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Repository stars | Main difference |
|---|---|---|---|
| ADLC Harness (this card) | no | 125 | |
| HarnessKit | no | 491 | an app for managing skills and configs of 17 agents; ADLC Harness defines the work loop and memory inside a plugin |
| Agent Harness | no | 313 | an agent scaffold; here a ready process from intent to PR with approvals |
README in short
The README explains the stage-by-stage workflow, a table of stages and skills, the memory layers and task exchange between lead and worker agents through a file mailbox.