claude

vibeforge

Vibeforge — a Claude Code governance methodology: multi-agent orchestration, discovery-driven lab scaffolding, and organic capitalization across projects.

1 on GitHubby mvlvgasyv0.4.1SEE LICENSE IN LICENSEupdated 12 days ago
/

Plugin vibeforge from marketplace vibeforge-marketplace in mvlvgasy/vibeforge-os. Commands read from the repo's manifest — 11 days ago.

This plugin runs code on your machine — its hooks fire automatically on events like tool use. The exact commands are listed below; read them before installing.

What it installs

Skills (37)

  • architecte

    Launches the Architect. Specs, ADRs, locked decisions. Never writes code. Runs after BA framing, before prompt-engineer.

  • audit-memory-age

    Scans the agents' MEMORY.md files (3 levels), computes age, detects amnesic agents (>14d without update despite invocations), and the R016 word ceiling (WARNING >=450, VIOLATION >=500). Report + recommendations.

  • business-analyst

    Launches the Business Analyst. Discovery, framing, PRD writing, stakeholders, success criteria. First agent invoked on a new project.

  • cloture-cadrage

    Closes a long framing phase (R007) - Lead synthesis, MEMORY update, releases routing. Use before switching to build or moving to another effort. Distinct from /cloture-session.

  • cloture-session

    End-of-session audit - registers, pending-blockers/eval, HANDOVER, agent MEMORY updates, promotion candidates. Triggered by SessionEnd hook or manually.

  • code-reviewer

    Launches the Code Reviewer. Double layer (superpowers + method patterns). Verdict OK/MINOR/BLOCKING. End of every build session before push.

  • dedup-registres

    Detects and merges similar BDR/learnings via semantic embeddings. Dedup phase of the consolidation cycle (R009). Invoke periodically (monthly) or when an audit detected a duplicate.

  • dream

    MEMORY consolidation (R009). 3 modes: auto (session-start hook >24h), manual (HITL), global (stack-wide, cross-lab). Strict MEMORY scope only — not the registers.

  • forge-worker

    Create a custom DW worker in agents/_workers/ (lightweight, inlined context, cheap model). To grow the reusable fleet. Callable only AFTER a session restart (registration happens at startup).

  • lab-architect

    Launches the Lab Architect. Creates a new lab: discovery, naming, complementary agents, SOUL/CLAUDE.md drafts, /new-lab invocation.

  • lab-factory

    Fabrique un lab standalone COMPILÉ depuis le métier du client — scan brownfield avant toute question, interview LAB_BRIEF, gate de complétude machine-vérifiée (check-brief.ps1), génération modulaire, puis compilation des agents/skills métier depuis les capacités justifiées. Remplace le duo "générer

  • lead-auto

    Launches the Lead in autonomous mode. Chains delegations without pause-validation. For already-framed projects where each step is predictable.

  • lead

    Launches the Lead (semi-auto mode). Multi-agent orchestration, project framing, synthesis. Default entry point for any complex request.

  • mass-code-review

    Parallel review of a large diff — 1 lightweight worker per module via Dynamic Workflow, adversarial verification of blockers, aggregation + global verdict. For changesets >5 files.

  • mass-memory-update

    Parallel update of the MEMORY.md files of amnesic agents (detected by /audit-memory-age) via Dynamic Workflow. Each agent re-reads its history and compresses its MEMORY (R016 <=500 words). Writes to distinct files = safe.

  • metrics-report

    Vibeforge metrics report from events.jsonl: top usage, dormant agents/skills, archive candidates. Focus --focus=memory-consolidation for R009 metrics.

  • new-lab-standalone

    Generate a STANDALONE lab (doctrine frame 16) — frozen method embedded (_method/ socle + vendored generic agents), zero plugin/sibling dependency, ready to hand to anyone. Via scripts/new-lab-standalone.ps1. This is THE lab-creation path in Vibeforge OS.

  • new-lab

    DEPRECATED (2026-07-04, frame 16) — use /new-lab-standalone instead. Vibeforge OS now generates STANDALONE labs only (frozen socle embedded, zero dependency). This legacy skill created plugin-dependent labs.

  • new-projet

    Create a project split deliverable/meta (doctrine 12). Deliverable in projets/<category>/<name>/, meta in <lab>/projets-meta/<name>/. A single session with cwd=lab.

  • parallel-cadrage

    Parallel challenge of a PRD draft via Dynamic Workflow - devil-advocate + architecte (feasibility) + reviewer-prd (coverage R001) simultaneously, then synthesis. Speeds up framing VALIDATION (not discovery).

  • promote-decision-to-bdr

    Detects committing strategic decisions in HANDOVER.md/journal.md/notes and proposes promotion to bdr.md (method root or lab). Active phase of BDR capitalization.

  • promote-learning

    Promotes a local learning to the global registers (generic reformulation, [promoted] marking, traceability). Invoked by traffic-controller after operator validation.

  • promote-pattern-to-eval

    Detects LLM hallucination/error patterns in learnings.md/journal.md and proposes promotion to eval.md. Active phase of Eval capitalization.

  • prompt-engineer

    Launches the Prompt Engineer. Compiles PRD+architecture+HANDOVER+rules into a build prompt. Audited by reviewer-prd (R001) before sending.

  • recover-cadrage

    Rebuilds a project's lost framing memory by re-reading a .jsonl session transcript, via deterministic condensation + Dynamic Workflow. Last-resort safety net when a framing session has "zero memory" after multiple compactions (R018). Token-efficient (condenses 10-15x before reading).

  • review-prd-coverage

    PRD coverage audit before build (R001). Compares a candidate prompt vs PRD+architecture+HANDOVER+rules. Decisive verdict: OK / NOT OK / REVIEW.

  • reviewer-prd

    Launches the Reviewer PRD. Mandatory audit before build (R001). READ-ONLY, clear-cut decision OK/NOT OK/REVIEW. Invoked by the lead before each build session.

  • save-cadrage

    Saves / updates the persistent framing memory (MEMORY-cadrage.md) of the current project. Externalizes the why, roadmap, decisions and key facts to disk to survive compactions (R018). Run on each structuring decision, after a daily, before a compaction, or at end of session.

  • scan-obsoletes

    Detects obsolete register entries (status=active but last_updated >180d + 0 recent citation). Archive phase of the consolidation cycle (doctrine frame 3). Invoke monthly.

  • session-analyzer

    Analysis of .jsonl transcripts (cost, tokens, cache, compactions). Method 1 (default) = deterministic script analyze-transcript.mjs (reliable, instant, free). Method 2 = DW + worker to interpret N sessions in parallel. Correct Opus pricing ($5/$25).

  • skill-curator

    Launches the Skill Curator. Validates proposed skills (5 checks). Auto-validation if 4 criteria pass, otherwise operator validation required.

  • strategie-contenu

    Launches the Content Strategist. Positioning, editorial calendar, narrative hooks, n8n/Make automations. Content, marketing, freelance labs.

  • traffic-controller

    Launches the Traffic Controller. Cross-lab scan, proposes promotions/demotions. Gated on ≥3 active projects. Suggested by /dream when candidate patterns are detected.

  • upgrade-socle

    Upgrades the METHOD layer of an existing standalone lab (_method/ socle, method agents/skills, hooks, CATALOG) WITHOUT touching its custom layer (domain agents/skills, register content). Via scripts/upgrade-socle.ps1.

  • use-cases

    Detects and prioritizes an organization's AI use cases BEFORE any project scoping (frame 18, step 0). Pain-point interviews, UC sheets, value / feasibility / risk / effort grid, prioritized portfolio, then launch of the chosen use case as a project (/new-projet). For a PO / Product Builder who must

  • ux

    Launches the UX Designer. Flows, textual wireframes, tone guide, persona, a11y. Upstream of or in parallel with the architect.

  • verify-completion

    Verifies an agent's completion claims against concrete evidence (diff, test output). Triggered by the Stop hook on .claude/verification-required.md or manually.

Subagents (15)

  • architecte

    Technical design - specs, ADRs, locked diagrams. Never writes code. Invoked after business-analyst, before prompt-engineer.

  • business-analyst

    Business need scoping, PRD writing, stakeholder discovery and success criteria. Invoked upstream of any work item, before architecte.

  • code-reviewer

    Reviews code before push: bugs, security, client patterns, PRD alignment. Two layers (superpowers + Vibeforge domain). Invoked at the end of a build session. Never modifies the code.

  • devil-advocate

    Adversarial critique (PRD/architecture/plan): unverified assumptions, blind spots, analyst bias. DOES NOT BLOCK. On explicit request only, never by default.

  • dream-validator

    Validates MEMORY proposals (apply/reject/defer). Auto-applies when there are >=3 sources and a clear scope. Invoked after each memory-proposal run.

  • jarvis

    The lab's butler. Full-spectrum steering assistant: lab status, data, one-off actions, awareness of everything going on. Not specialized. Direct @jarvis conversation preferred. Never replaces the lead for orchestrating a workstream.

  • lab-architect

    Creates a new STANDALONE lab (frame 16) — domain discovery (mandatory floor), naming, complementary agents, SOUL/CLAUDE.md drafts, /new-lab-standalone invocation. Invoked by the lead when you want a new domain.

  • lead

    The single orchestrator. Delegates to specialists via the Task tool. Never writes code. 3 modes - /lead-auto (autonomous), /lead (semi-auto, default), direct slash command.

  • prompt-engineer

    Compiles PRD + architecture + addendum + HANDOVER + rules into a Claude Code session prompt. Invoked before each build, audited by reviewer-prd (R001) afterward.

  • reviewer-prd

    Audits PRD coverage before any build session (R001). Delivers a clear OK / NOT OK / REVIEW verdict. READ-ONLY — flags issues, does not fix them.

  • skill-curator

    Validates skills proposed by agents (5 checks: duplicates, overlap, granularity, format, relevance). Decides approve/merge/revise/reject. Invoked before any write into skills/.

  • strategie-contenu

    Content strategy and marketing automation: editorial calendars, narrative hooks, n8n/Make workflows. For content, marketing, freelance labs.

  • traffic-controller

    Cross-lab scan, proposes promotions (learnings → global rules) and demotions. Never promotes without operator validation. Active at ≥3 active projects.

  • ux

    User journeys, textual wireframes, tone guide, personas, a11y. Never codes. Invoked before or in parallel with the architect on user-facing apps.

  • verifier

    Adversarial goal-backward verification AFTER a build. Starts from the PRD success criteria, REQUIREMENTS and must_haves, never from the builder's report. Verdicts VERIFIED / FAILED / UNCERTAIN. Reads and runs proofs, never modifies the deliverable.

Hooks (8)

  • SessionStart

    powershell.exe -ExecutionPolicy Bypass -File "${CLAUDE_PLUGIN_ROOT}\hooks\session-start.ps1"

  • Stop

    powershell.exe -ExecutionPolicy Bypass -File "${CLAUDE_PLUGIN_ROOT}\hooks\stop.ps1"

  • SessionEnd

    powershell.exe -ExecutionPolicy Bypass -File "${CLAUDE_PLUGIN_ROOT}\hooks\session-end.ps1"

  • PreToolUse

    powershell.exe -ExecutionPolicy Bypass -File "${CLAUDE_PLUGIN_ROOT}\hooks\pre-tool-use.ps1"

  • PreToolUse

    powershell.exe -ExecutionPolicy Bypass -File "${CLAUDE_PLUGIN_ROOT}\hooks\config-protection.ps1"

  • PreToolUse

    powershell.exe -ExecutionPolicy Bypass -File "${CLAUDE_PLUGIN_ROOT}\hooks\fact-force.ps1"

  • PostToolUse

    powershell.exe -ExecutionPolicy Bypass -File "${CLAUDE_PLUGIN_ROOT}\hooks\post-tool-use.ps1"

  • PreCompact

    powershell.exe -ExecutionPolicy Bypass -File "${CLAUDE_PLUGIN_ROOT}\hooks\pre-compact.ps1"

View source on GitHub

Similar plugins

Tenant-isolated AI teams that plan work, coordinate tasks, recall approved context, and return evidence-backed results

6 skills · 6 commands · 4 agents · 1 MCP server

ruvnet/ruflo 81,589

Build powerful, stateful CLI interfaces for any GUI application using the cli-anything harness methodology.

5 commands

HKUDS/CLI-Anything 58,665

data-agent-kit

claudecodex

This plugin provides a specialized suite of skills for data engineers and database practitioners working on Google Cloud. It acts as an expert assistant, allowing you to use natural language prompts in your preferred coding agent to architect complex data pipelines, transform data with dbt, write Spark and BigQuery SQL notebooks, and orchestrate end-to-end workflows across GCP's data ecosystem.

37 skills

anthropics/claude-plugins-official 47,076

Backend API design, GraphQL architecture, workflow orchestration with Temporal, and test-driven backend development

9 skills · 1 command · 8 agents

wshobson/agents 42,461

Claude Code native multi-agent orchestration with intelligent model routing across agents and skills. Zero learning curve. Maximum power.

85 skills · 20 commands · 19 agents · 1 MCP server · hooks

Yeachan-Heo/oh-my-claudecode 40,800

Web scraping, Google search, structured data extraction, and MCP server integration powered by Bright Data. Includes 7 skills: scrape any webpage as markdown (with bot detection/CAPTCHA bypass), search Google with structured JSON results, extract data from 40+ websites (Amazon, LinkedIn, Instagram, TikTok, YouTube, and more), orchestrate Bright Data's 60+ MCP tools, built-in best practices for Web Unlocker, SERP API, Web Scraper API, and Browser API, Python SDK best practices for the brightda...

21 skills

anthropics/knowledge-work-plugins 35,601

Maintain vibeforge?

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 pluginslisted