Specloop — Testing skill for Claude Code
Spec-driven loop engineering: a repeatable construct for driving coding agents through numbered phase specs, with an enforced structure and mandatory status handoffs.
How to install Specloop
This entry records only its repository, not the path inside it, so there is no
exact command to give. Open khangtoh/specloop and copy the folder into
~/.claude/skills/, or the file into ~/.claude/agents/.
What Specloop does
Spec-driven loop engineering: a repeatable construct for driving coding agents through numbered phase specs, with an enforced structure and mandatory status handoffs. Installs as a bun CLI, a Claude Code plugin, and a Codex plugin.
Alternatives in Testing
- Langgraphjs — by langchain-ai - Offers comprehensive build and test commands with detailed TypeScript style guidelines, laye 2.7k ★
- AI Contribution Skills (Engram) — branch-pr: clean branch and PR workflow 2.5k ★
- Loki Spec Status — Check whether the spec has drifted from its lock using Loki's living-spec drift detection, and summarize the r 1.1k ★
README
specloop
**Spec-driven loop engineering** — a repeatable, enforceable construct for driving agents through a project, extracted so it's consistent across every repo you use it in.
A goal is decomposed into **numbered phase specs**, each a flat checklist of atomic tasks. An **agent loop** executes the next unchecked box, verifies it, and hands off in a **mandatory report format**. A **bun validator** enforces the directory structure so it can't silently drift, and a **goal-completion-check** prompt audits whether a goal is actually met — by tracing it through requirements → specs → checkboxes, not vibes.
It installs as a **Claude Code plugin**, a **Codex plugin**, and a standalone **bun CLI**. The same `template/` directory is what the CLI scaffolds, what the validator checks against, and what you inspect to learn the method.
What you get
| Piece | What it is |
|---|---|
spec/README.md |
Phase index — the goal, the # | File | Purpose | Status | Blocking dependency table, and non-goals. |
spec/BACKLOG.md |
The ranked work order — one line per phase; the loop takes the top. Stores order only; done-state is derived. prio-spec edits it. |
spec/NN-title.md |
A phase — Goal: + Depends on: header, flat - [ ] / - [x] atomic tasks, and a dated Findings / Results log. |
spec/spec-summary-status.md |
The mandatory Spec Summary/Status handoff every agent must emit. |
spec/goal-completion-check.md |
A reusable prompt tracing a goal → requirements → specs → checkboxes. |
spec/agent-session-ledger.md |
A dated narrative log of what each session did and left running. |
AGENTS.md |
Binds all agents in the repo to the reporting standard and the loop. |
.specloop.json |
Validator config. |
Architecture: three layers
specloop's constructs group into three layers, each answering one question. The validator (`specloop check`) sits underneath all three, keeping every derived fact (counts, status, done-state) honest
Related Skills
Implement Phase
Autonomously execute a whole roadmap phase - resolve features to specs, create missing ones, then loop impleme
Piv
Self-recursive development loop with real auth testing, live DB research, 100% test pass requirements, and MAN
Cfn Arch
SPARC Architecture phase. Define component boundaries, interface contracts, integration points, DRY reuse BEFO
Q Status
Check Spec-Kit-Plus workflow status with Quality Gate grades - shows phase progress, validation reports, and n
Validate Phase
Run a phase's acceptance gates and the closure test; flip status to done only on a clean pass
Agent Tools Claude Code
Agent Tools is a Claude Code plugin marketplace for engineering tasks, automated testing, code quality, browse
Related Agents
Hatch3r Handoff Loader
Session-start agent that surfaces active handoff documents from .hatch3r/handoffs/active/. Use at the beginnin
Subagent Docs Analyst
Documentation Specialist. MANDATORY at the end of EVERY implementation (Vibe Coding or Subagents mode). Manage
Socratic Questioner
Mandatory pre-implementation interrogation. Use before any non-trivial coding task to surface assumptions, fin