By jeanfbrito
Multi-agent orchestration conventions: orchestrator mode, 10 subagent roles, slash commands, hooks.
Dispatch the `planner` subagent to handle `$ARGUMENTS`. This is the front door for any non-trivial task. The tier flag controls pipeline depth — and therefore cost.
Append a decision blocker to .localdev/workflow/blockers.md in canonical format and halt the current task until the user resolves it
Write a cross-session handoff for the current task, so the next session can resume with full context
Scaffold the Agentic Workflow Framework files for the current project (.localdev/workflow/ + docs/KNOWN_ISSUES.md)
Document a persistent platform or dependency constraint in docs/KNOWN_ISSUES.md — the kind of thing the next developer or agent would re-discover otherwise
Escalation agent dispatched after 2 failed attempts at the same problem. Diagnoses the root constraint (not the symptom) and redesigns the approach. Called to think, not to code. Use when the Planner's pipeline has stalled twice.
One small, well-defined task — a single rename, stub, typo, or config tweak. Unambiguous, but the change must be assembled, not just repeated across files. Use for a single scoped edit; for the SAME edit repeated across 5+ sites use builder-trivial instead.
Exception-tier implementation — opus for strategy, sonnet for attacks; implementation defaults to builder-fast. Dispatch when a sonnet attempt failed, or when the code itself demands strategy-grade reasoning no brief can pre-decide (novel algorithms, subtle concurrency). Serialize by file (no two smart Builders on the same file simultaneously).
Repetitive bulk edits — the SAME change applied across many sites (5+ files/entries): mass renames, bulk i18n/config additions, stub generation. One fully-specified transform, zero per-site decisions. Cheapest tier (haiku). Pick this only when the task is "apply X to N places", not "build one small thing".
Fast codebase search specialist. Finds files by pattern, traces call chains, maps patterns across the tree. Read-only and parallel-safe. Use for "where is X?", "what calls Y?", "which files match Z?"
Use automatically for non-trivial engineering work: multi-step tasks, changes across 2+ files, risky or cross-cutting refactors, debugging with unknown root cause, multi-session work, structured task ledgers, blockers, handoffs, known issues, and verification planning. Drives Orchestrator dispatch-by-default with planner/finder/researcher/builder/reviewer/tester/auditor/watcher subagents.
Use automatically as global engineering behavior guidance: finding root causes over temporary fixes, understand-before-changing, reference-first porting, demanding elegant solutions, cautious treatment of review-bot feedback, evidence-based verification, never committing/pushing without explicit ask, careful customer-facing wording, and code-intelligence tool selection.
Modifies files
Hook triggers on file write and edit operations
Uses power tools
Uses Bash, Write, or Edit tools
Own this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimOwn this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimBased on adoption, maintenance, documentation, and repository signals. Not a security audit or endorsement.
No model invocation
Executes directly as bash, bypassing the AI model
No model invocation
Executes directly as bash, bypassing the AI model
Lightweight multi-agent orchestration conventions for Claude Code.
~/.claude/AGENTIC.md) -- the full framework spec imported into every session via CLAUDE.md~/.claude/agents/: planner, auditor, reviewer, builder-smart, builder-fast, builder-trivial, finder, researcher, tester, watcher~/.claude/commands/: /agentic, /init-agentic, /handoff, /blocker, /known-issue, /qq~/.claude/hooks/orchestrator.sh) -- fires on UserPromptSubmit to prevent Orchestrator driftsettings.json hook entries -- SessionStart (blocker/handoff scanner) + UserPromptSubmit (orchestrator reinforcement)CLAUDE.md import line -- @AGENTIC.md so the spec loads globally in every Claude Code session~/.claude/skills/: agentic-workflow and personal-engineering-rules~/.claude/settings.json: Write/Edit for .localdev/workflow/**, .localdev/workflow/handoffs/**, and docs/KNOWN_ISSUES.mdThis framework treats three external tools as first-class infrastructure:
These are not bundled. Install separately. Fail-loud rule: when one of their MCP calls errors, the orchestrator and every subagent report the exact tool + verbatim error to the user instead of silently degrading to raw grep / web / training-data recall — silent fallback is how hallucinations ship.
Option 0: Claude Code plugin marketplace (no cloning)
/plugin add jeanfbrito/agentic-workflow-framework
Claude Code downloads, installs, and activates the framework automatically. Hooks, slash commands, agents, and skills are live immediately — no shell step required.
The plugin sandbox cannot edit ~/.claude/CLAUDE.md or settings.json, so the plugin closes those two gaps with its own hooks:
AGENTIC.md into every session's context (it no-ops when the shell installer's @AGENTIC.md import is present, so nothing loads twice)..localdev/workflow/** and docs/KNOWN_ISSUES.md, mirroring the permission globs install.sh grants. All other paths follow the normal permission flow.Both paths land on the same conventions but wire them up differently: the plugin path relies on its own hooks (context injection, orchestrator reinforcement, permission auto-allow) since it cannot touch ~/.claude/CLAUDE.md or settings.json; the shell path instead writes the @AGENTIC.md import into CLAUDE.md and the permission globs directly into settings.json. Same end behavior via different mechanisms.
Option 1: clone and run
git clone https://github.com/jeanfbrito/agentic-workflow-framework.git ~/agentic-workflow-framework
cd ~/agentic-workflow-framework
./install.sh
Pass --yes / -y to skip the confirmation prompt. Pass --link to symlink instead of copy (dev mode — edits to the repo are reflected immediately).
The agent-driven self-installer has been retired (its embedded file copies drifted from the source of truth). Use the plugin marketplace path or the shell installer above.
./uninstall.sh
Removes the framework files, installed skills, and permission globs. By default this is a surgical strip: it removes only the framework's hook entries, permission globs, and @AGENTIC.md import, preserving anything else you've added to CLAUDE.md or settings.json since installing. Pass --restore-backup to instead wholesale-restore CLAUDE.md and settings.json from the pre-install backup snapshot, discarding all edits made since install. Does not remove the ~/.claude/agents/, ~/.claude/commands/, ~/.claude/hooks/, or ~/.claude/skills/ directories themselves.
Doctrine: opus is the general, sonnet the soldiers, haiku the scouts.
Multi-OS Desktop App Testing with GPU passthrough — VLM-driven UI verification.
npx claudepluginhub jeanfbrito/agentic-workflow-framework --plugin agentic-workflow-frameworkUltra-compressed communication mode. Cuts 65% of output tokens (measured) while keeping full technical accuracy by speaking like a caveman.
Multi-model consensus engine integrating OpenAI Codex CLI, Gemini CLI, and Claude CLI for collaborative code review and problem-solving.
Comprehensive UI/UX design plugin for mobile (iOS, Android, React Native) and web applications with design systems, accessibility, and modern patterns
Unified capability management center for Skills, Agents, and Commands.
Memory compression system for Claude Code - persist context across sessions
Standalone image generation plugin using Nano Banana MCP server. Generates and edits images, icons, diagrams, patterns, and visual assets via Gemini image models. No Gemini CLI dependency required.