agent-manager-skill

Estrelas10

Forks2

Atualizado10 de março de 2026 às 08:52

Manage multiple local CLI agents via tmux sessions (start/stop/monitor/assign) with cron-friendly scheduling. Includes agent performance optimization workflows.

Instalação

Instalar com Codex ou Claude Copie este prompt, cole no Codex, Claude ou outro assistente e deixe que ele revise a página da skill e instale para você.

Fonte

bugrabilge

bugrabilge/bilge-development-kit

Abrir repositório GitHub Ver repositórios do creator

Download

Ocupações relacionadasSOC

Baseado na classificação ocupacional SOC

Administradores de redes e sistemas de computadorInformática e Matemática·SOC 15-1244

SKILL.md

readonly

name	agent-manager-skill
description	Manage multiple local CLI agents via tmux sessions (start/stop/monitor/assign) with cron-friendly scheduling. Includes agent performance optimization workflows.
risk	unknown
source	community

Agent Manager Skill

When to use

Use this skill when you need to:

run multiple local CLI agents in parallel (separate tmux sessions)
start/stop agents and tail their logs
assign tasks to agents and monitor output
schedule recurring agent work (cron)

Prerequisites

Install agent-manager-skill in your workspace:

git clone https://github.com/fractalmind-ai/agent-manager-skill.git

Common commands

python3 agent-manager/scripts/main.py doctor
python3 agent-manager/scripts/main.py list
python3 agent-manager/scripts/main.py start EMP_0001
python3 agent-manager/scripts/main.py monitor EMP_0001 --follow
python3 agent-manager/scripts/main.py assign EMP_0002 <<'EOF'
Follow teams/fractalmind-ai-maintenance.md Workflow
EOF

Notes

Requires tmux and python3.
Agents are configured under an agents/ directory (see the repo for examples).

Agent Performance Optimization Workflow

Systematic improvement of existing agents through performance analysis, prompt engineering, and continuous iteration.

Phase 1: Performance Analysis and Baseline

Gather Performance Data

Collect metrics including:

Task completion rate (successful vs failed tasks)
Response accuracy and factual correctness
Tool usage efficiency (correct tools, call frequency)
Average response time and token consumption
User satisfaction indicators (corrections, retries)
Hallucination incidents and error patterns

User Feedback Pattern Analysis

Identify recurring patterns:

Correction patterns: Where users consistently modify outputs
Clarification requests: Common areas of ambiguity
Task abandonment: Points where users give up
Follow-up questions: Indicators of incomplete responses
Positive feedback: Successful patterns to preserve

Failure Mode Classification

Categorize failures by root cause:

Instruction misunderstanding, output format errors
Context loss in long conversations
Tool misuse or inefficient tool selection
Constraint violations, edge case handling

Baseline Performance Report

- Task Success Rate: [X%]
- Average Corrections per Task: [Y]
- Tool Call Efficiency: [Z%]
- User Satisfaction Score: [1-10]
- Average Response Latency: [Xms]
- Token Efficiency Ratio: [X:Y]

Phase 2: Prompt Engineering Improvements

Chain-of-Thought Enhancement: Add explicit reasoning steps and self-verification checkpoints
Few-Shot Example Optimization: Curate diverse examples covering common use cases and edge cases
Role Definition Refinement: Clear mission, expertise domains, behavioral traits, constraints
Constitutional AI Integration: Self-correction with critique-and-revise loops
Output Format Tuning: Structured templates, dynamic formatting, progressive disclosure

Phase 3: Testing and Validation

A/B Testing Framework

Minimum sample size: 100 tasks per variant
Confidence level: 95% (p < 0.05)
Compare: success rate, speed, token usage
Use blind human review + automated scoring

Evaluation Metrics

Task-Level: Completion rate, correctness, efficiency, tool appropriateness
Quality: Hallucination rate, consistency, format compliance, safety
Performance: Latency, token consumption, cost per task

Phase 4: Version Control and Deployment

Version Management

Format: agent-name-v[MAJOR].[MINOR].[PATCH]
MAJOR: Significant capability changes
MINOR: Prompt improvements, new examples
PATCH: Bug fixes, minor adjustments

Staged Rollout

Alpha (5% traffic) -> Beta (20%) -> Canary (20-50-100%) -> Full deployment
7-day monitoring period after full deployment

Rollback Triggers

Success rate drops >10% from baseline
Critical errors increase >5%
Cost per task increases >20%
Safety violations detected

Success Criteria

Task success rate improves by >=15%
User corrections decrease by >=25%
No increase in safety violations
Response time remains within 10% of baseline

Continuous Improvement Cycle

Weekly: Monitor metrics and collect feedback
Monthly: Analyze patterns and plan improvements
Quarterly: Major version updates with new capabilities

Mais deste repositório

mesmo repositório

audit

bugrabilge/bilge-development-kit

Full codebase audit: architecture health, dependencies, complexity, tech debt. Use when: "audit", "health check", "codebase analysis", "tech debt", "architecture review", "dependency check", "code quality"

2026-04-0410

brainstorm

bugrabilge/bilge-development-kit

Structured brainstorming with multiple approaches and tradeoffs. Use when: "brainstorm", "explore options", "what are my options", "compare approaches", "pros and cons", "which is better", "how should I", "alternatives"

2026-04-0410

build-fix

bugrabilge/bilge-development-kit

Build error diagnosis and resolution. Use when: "build failed", "compile error", "bundler error", "fix build", "npm error", "type error", "won't compile"

2026-04-0410

create

bugrabilge/bilge-development-kit

Scaffold a new application with interactive dialogue. Use when: "create", "build me", "make an app", "new project", "scaffold", "generate app", "start new"

2026-04-0410

create-skill

bugrabilge/bilge-development-kit

Generate a new custom skill from natural language description. Use when: "create skill", "new skill", "make a skill", "add command", "custom workflow", "skill template", "generate skill"

2026-04-0410

debug

bugrabilge/bilge-development-kit

Systematic debugging with hypothesis-driven investigation. Use when: "debug", "not working", "bug", "broken", "error", "unexpected behavior", "fix issue", "investigate"

2026-04-0410

name	agent-manager-skill
description	Manage multiple local CLI agents via tmux sessions (start/stop/monitor/assign) with cron-friendly scheduling. Includes agent performance optimization workflows.
risk	unknown
source	community

Agent Manager Skill

When to use

Use this skill when you need to:

run multiple local CLI agents in parallel (separate tmux sessions)
start/stop agents and tail their logs
assign tasks to agents and monitor output
schedule recurring agent work (cron)

Prerequisites

Install agent-manager-skill in your workspace:

git clone https://github.com/fractalmind-ai/agent-manager-skill.git

Common commands

python3 agent-manager/scripts/main.py doctor
python3 agent-manager/scripts/main.py list
python3 agent-manager/scripts/main.py start EMP_0001
python3 agent-manager/scripts/main.py monitor EMP_0001 --follow
python3 agent-manager/scripts/main.py assign EMP_0002 <<'EOF'
Follow teams/fractalmind-ai-maintenance.md Workflow
EOF

Notes

Requires tmux and python3.
Agents are configured under an agents/ directory (see the repo for examples).

Agent Performance Optimization Workflow

Systematic improvement of existing agents through performance analysis, prompt engineering, and continuous iteration.

Phase 1: Performance Analysis and Baseline

Gather Performance Data

Collect metrics including:

Task completion rate (successful vs failed tasks)
Response accuracy and factual correctness
Tool usage efficiency (correct tools, call frequency)
Average response time and token consumption
User satisfaction indicators (corrections, retries)
Hallucination incidents and error patterns

User Feedback Pattern Analysis

Identify recurring patterns:

Correction patterns: Where users consistently modify outputs
Clarification requests: Common areas of ambiguity
Task abandonment: Points where users give up
Follow-up questions: Indicators of incomplete responses
Positive feedback: Successful patterns to preserve

Failure Mode Classification

Categorize failures by root cause:

Instruction misunderstanding, output format errors
Context loss in long conversations
Tool misuse or inefficient tool selection
Constraint violations, edge case handling

Baseline Performance Report

- Task Success Rate: [X%]
- Average Corrections per Task: [Y]
- Tool Call Efficiency: [Z%]
- User Satisfaction Score: [1-10]
- Average Response Latency: [Xms]
- Token Efficiency Ratio: [X:Y]

Phase 2: Prompt Engineering Improvements

Chain-of-Thought Enhancement: Add explicit reasoning steps and self-verification checkpoints
Few-Shot Example Optimization: Curate diverse examples covering common use cases and edge cases
Role Definition Refinement: Clear mission, expertise domains, behavioral traits, constraints
Constitutional AI Integration: Self-correction with critique-and-revise loops
Output Format Tuning: Structured templates, dynamic formatting, progressive disclosure

Phase 3: Testing and Validation

A/B Testing Framework

Minimum sample size: 100 tasks per variant
Confidence level: 95% (p < 0.05)
Compare: success rate, speed, token usage
Use blind human review + automated scoring

Evaluation Metrics

Task-Level: Completion rate, correctness, efficiency, tool appropriateness
Quality: Hallucination rate, consistency, format compliance, safety
Performance: Latency, token consumption, cost per task

Phase 4: Version Control and Deployment

Version Management

Format: agent-name-v[MAJOR].[MINOR].[PATCH]
MAJOR: Significant capability changes
MINOR: Prompt improvements, new examples
PATCH: Bug fixes, minor adjustments

Staged Rollout

Alpha (5% traffic) -> Beta (20%) -> Canary (20-50-100%) -> Full deployment
7-day monitoring period after full deployment

Rollback Triggers

Success rate drops >10% from baseline
Critical errors increase >5%
Cost per task increases >20%
Safety violations detected

Success Criteria

Task success rate improves by >=15%
User corrections decrease by >=25%
No increase in safety violations
Response time remains within 10% of baseline

Continuous Improvement Cycle

Weekly: Monitor metrics and collect feedback
Monthly: Analyze patterns and plan improvements
Quarterly: Major version updates with new capabilities