claude

agent-loop

Agent-centric development cycle. Guide (agent-config), Verify (grill-me, peer-review), Solve (ralph-loop, issue-loop) — plus the ralph-audit command.

1 on GitHubby Victor Hueniv1.6.1MITupdated 6 days ago
/

Plugin agent-loop from marketplace homemade-claude-kit in VictorHueni/homemade-claude-kit. Commands read from the repo's manifest — 6 days ago.

What it installs

Skills (6)

  • agent-config

    Scaffold, review, and improve CLAUDE.md and AGENTS.md files for any project. Applies behavioral posture guidelines, token budget discipline, progressive disclosure, and docs-index wiring so agents navigate the repo without stale inline path listings. Three modes: scaffold (minimal config from projec

  • agent-grill-me

    Interactive Socratic stress-test of a PRD or implementation plan: one focused question at a time, project domain vocabulary checked against docs/domain/glossary.md, decisions crystallised into ADRs via arch-adr. Triggers on: grill me on this PRD, grill me on this plan, stress-test this with me, ques

  • agent-handoff

    Hand off an overloaded interactive session's context to a fresh session via a validated markdown handoff document. Two modes: create (write the handoff at var/handoffs/, end with a paste-ready resume prompt) and resume (load a handoff in a fresh session, check repo drift, re-verify inherited claims,

  • agent-issue-loop

    Work a GitHub-backend issue backlog as a coding agent: validate an issue's readiness (or sweep the ready queue for drift), find and merge duplicate issues, and take a ready-for-agent issue through implement → verify → PR. Triggers on: take issue #N, work the next ready issue, next issue, validate is

  • agent-peer-review

    Critically review PRDs and implementation plans to identify gaps, blind spots, contradictions, and delivery risks before coding begins. Use when asked to review product docs, technical specs, implementation plans, migration/deletion plans, acceptance criteria, or rollout strategy; produce ranked fin

  • agent-ralph-loop

    Execute an implementation plan autonomously using the Ralph Loop protocol. Iterates through increments one at a time: implement, test, commit, repeat. Use when asked to 'run the ralph loop', 'execute this plan', or 'start autonomous execution'.

Slash commands (1)

  • /agent-loop:ralph-audit

    Audit an execution plan against Ralph Loop requirements and emit the bash command to run it

View source on GitHub

More from VictorHueni/homemade-claude-kit

All 11 plugins →

strategy

claude

Business strategy & discovery. Vision, personas, canvases, capability map, value streams, objectives, processes, quantitative models, competitive landscape — plus discovery ideation, research, and workshops.

12 skills

1

dev-flow

claude

Developer workflow. Git enforcement stack, conventional commits, PRs, worktrees, release automation, stack guides, onboarding — plus the branch-cleanup-audit command.

7 skills · 1 command · 2 MCP servers

1

Architecture. ADRs, research notes, service/CLI contracts, Structurizr/C4, PlantUML/UML, and arc42 narrative.

9 skills

1

Delivery communication. HTML slide decks, artefact visualisations, stakeholder release notes, and the design system that themes them.

4 skills · 1 MCP server

1

kit-core

claude

Kit core — always enabled. The metamodel keystone skill (spine, YAML type registry, frontmatter schema, open-items contract, Audit/Scaffold/Migrate modes) plus the cross-cutting operators: open-items ledger, provenance, toolkit doctor; and the tripwire hooks (commands with a known trap are blocked or annotated; a first write to auto-memory shows the where-does-this-belong checklist).

4 skills · hooks

1

Product specs & planning. FBS, PRDs, use cases, delivery roadmap, and implementation plans.

5 skills

1

Maintain agent-loop?

This page updates itself from your repo. Add the badge to your README, or get it Featured to pin it above the organic results with a followed link to your site.

claude & codex plugins11 plugins