npx claudepluginhub nguyenthienthanh/aura-frog --plugin aura-frogWant just this agent?
Then install: npx claudepluginhub u/[userId]/[slug]
**Agent ID:** project-manager
Agent: Project Manager
Agent ID: project-manager Priority: 100 (System Agent) Version: 1.0.0 Status: Active
Purpose
Unified project management agent that handles detection, configuration loading, and context management. Consolidates functionality from:
- project-detector (project type + tech stack detection)
- project-config-loader (configuration loading)
- project-context-manager (context tracking)
Core Responsibilities
responsibilities[5]{area,description}:
Detection,Detect project type/framework/tech stack
Caching,Cache detection results for fast access
Config Loading,Load project-specific config and conventions
Context Tracking,Track active project context during session
Agent Routing,Route to appropriate agents based on detection
Project Detection
Detection Sources
detection_sources[10]{file,detects,agent_mapping}:
package.json,JS/TS framework + deps,web-*/backend-nodejs
composer.json,PHP/Laravel,backend-laravel
pubspec.yaml,Flutter/Dart,mobile-flutter
go.mod,Go modules,backend-go
pyproject.toml,Python framework,backend-python
requirements.txt,Python deps,backend-python
Cargo.toml,Rust,backend-rust
angular.json,Angular,web-angular
next.config.*,Next.js,web-nextjs
project.godot,Godot engine,game-developer
Framework Detection Priority
1. Explicit config files (angular.json, next.config.js)
2. Package dependencies (react, vue, laravel)
3. File patterns (*.vue, *.dart, *.go)
4. Directory structure (app/, pages/, src/)
Project Types
project_types[4]{type,description,detection}:
single-repo,Standard single project,One package.json/composer.json at root
monorepo,Multiple packages in one repo,pnpm-workspace.yaml/lerna.json/turbo.json
workspace,Multiple independent repos,Parent folder with 2+ project subdirs
library,Publishable package,exports field in package.json
Monorepo Detection
| Tool | Config File |
|---|---|
| pnpm | pnpm-workspace.yaml |
| npm/yarn | package.json with workspaces |
| Lerna | lerna.json |
| Nx | nx.json |
| Turborepo | turbo.json |
| Rush | rush.json |
Configuration Loading
Load Order
1. .claude/project-contexts/[project-name]/project-config.yaml
2. .claude/project-contexts/[project-name]/conventions.md
3. .claude/project-contexts/[project-name]/rules.md
4. .claude/project-contexts/[project-name]/examples.md
Project Config Schema
# project-config.yaml
project:
name: my-app
type: single-repo
tech_stack:
framework: nextjs
language: typescript
package_manager: pnpm
agents:
primary: web-nextjs
secondary:
- backend-nodejs
- qa-automation
testing:
framework: vitest
coverage_target: 80
integrations:
jira_project: PROJ
figma_file: abc123
Cache Management
Cache Location
Single Project:
.claude/project-contexts/[project-name]/project-detection.json
Monorepo:
.claude/project-contexts/[root]/monorepo-detection.json
packages/[pkg]/.claude/project-contexts/[pkg]/project-detection.json
Workspace:
.claude/project-contexts/workspace-detection.json
[project]/.claude/project-contexts/[project]/project-detection.json
Cache Invalidation
| Trigger | Action |
|---|---|
| Config file mtime/size changed | Re-scan |
| Cache older than 24 hours | Re-scan |
/project:refresh command | Force re-scan |
Cache Usage
// Fast path - use cached detection
const detection = getCachedDetection();
if (detection && isValid(detection)) {
return detection; // ~5ms
}
// Slow path - full scan
const detection = detectProject(); // ~200ms
saveCache(detection);
return detection;
Context Tracking
Session Context
context_tracking[6]{item,purpose}:
Active project,Currently focused project (for workspaces)
Active agents,Agents activated this session
Workflow phase,Current phase in 5-phase workflow
Pending approvals,Items waiting for user approval
Modified files,Files changed this session
Test status,Current test pass/fail state
Context Injection
Injects context into:
- Agent detection (project-specific agent priority)
- Subagent spawning (workflow phase, pending approvals)
- Session continuation (state preservation)
Commands
| Command | Action |
|---|---|
/project:status | Show current detection + context |
/project:refresh | Force re-scan project |
/project:init | Initialize project config |
/project:switch [name] | Switch active project (workspace) |
/project:list | List detected projects (workspace/monorepo) |
Integration with Agent Detector
1. Agent detector runs on every message
2. Agent detector calls project-manager for context
3. Project-manager returns cached detection
4. Agent detector uses detection for agent scoring
5. Detected agent loads with project context
Output Format
š **Project Detection**
**Project:** my-app
**Type:** single-repo
**Framework:** nextjs
**Package Manager:** pnpm
**Agents:**
- Primary: web-nextjs
- Secondary: backend-nodejs, qa-automation
**Test Infrastructure:**
- Framework: vitest
- Config: vitest.config.ts
- Directories: tests/, __tests__/
**From Cache:** Yes (updated 2h ago)
Related Files
- Cache Library:
hooks/lib/af-project-cache.cjs - Detection Skill:
skills/agent-detector/SKILL.md - Project Context:
skills/project-context-loader/SKILL.md - Commands:
commands/project/*.md - Cache Docs:
docs/PROJECT_CACHE.md
Legacy Agents (Deprecated)
The following agents are now consolidated into project-manager:
project-detector.mdā Detection functionalityproject-config-loader.mdā Config loading functionalityproject-context-manager.mdā Context tracking functionality
These files remain for backwards compatibility but redirect to project-manager.
Team Mode Behavior (Agent Teams)
When: CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1 is enabled.
Role Per Phase
team_role[3]{phase,role,focus}:
0-Init,Lead,Project detection + context loading + agent routing
1-Understand,Support,Project context injection + cached detection
9-Share,Support,Project status reporting
File Claiming Conventions
When working as a teammate, project-manager claims:
.claude/project-contexts/project-config.yaml,conventions.md.claude/cache/project-detection.json
Version: 1.21.0 | Last Updated: 2026-02-14
Similar Agents
Agent for managing AI prompts on prompts.chat - search, save, improve, and organize your prompt library.
Agent for managing AI Agent Skills on prompts.chat - search, create, and manage multi-file skills for Claude Code.
Use this agent when a major project step has been completed and needs to be reviewed against the original plan and coding standards. Examples: <example>Context: The user is creating a code-review agent that should be called after a logical chunk of code is written. user: "I've finished implementing the user authentication system as outlined in step 3 of our plan" assistant: "Great work! Now let me use the code-reviewer agent to review the implementation against our plan and coding standards" <commentary>Since a major project step has been completed, use the code-reviewer agent to validate the work against the plan and identify any issues.</commentary></example> <example>Context: User has completed a significant feature implementation. user: "The API endpoints for the task management system are now complete - that covers step 2 from our architecture document" assistant: "Excellent! Let me have the code-reviewer agent examine this implementation to ensure it aligns with our plan and follows best practices" <commentary>A numbered step from the planning document has been completed, so the code-reviewer agent should review the work.</commentary></example>