Run any Skill in Manus with one click

juliaz-agent-improve

Systematic improvement of Julia's agents through performance analysis, prompt engineering, and validation. Trigger when optimizing any agent's behavior — orchestrator, frontend Julia, OpenClaw relay, or any sub-agent. Also trigger for: 'improve agent', 'agent not working well', 'optimize prompt', 'tool not being used correctly', 'Julia gives wrong answers', or any agent quality issue.

Run Skill in Manus

Overview

Install command

npx skills add https://github.com/abzhaw/juliaz_agents --skill juliaz-agent-improve

Copy and paste this command into Claude Code to install the skill

Source

abzhaw/juliaz_agents

Stars0

Forks0

UpdatedMarch 8, 2026 at 21:07

SKILL.md

readonly

name	juliaz-agent-improve
description	Systematic improvement of Julia's agents through performance analysis, prompt engineering, and validation. Trigger when optimizing any agent's behavior — orchestrator, frontend Julia, OpenClaw relay, or any sub-agent. Also trigger for: 'improve agent', 'agent not working well', 'optimize prompt', 'tool not being used correctly', 'Julia gives wrong answers', or any agent quality issue.

juliaz-agent-improve — Agent Performance Optimization

Systematic improvement of agents in the juliaz_agents system. Adapted from agent-orchestration-improve-agent.

Use this skill when

Improving an existing agent's performance or reliability (orchestrator, frontend Julia, OpenClaw)
Analyzing why an agent misuses tools, gives wrong answers, or fails silently
Optimizing system prompts for better tool selection
Validating agent changes before deployment

Do not use this skill when

Building a brand-new agent from scratch (use juliaz-agent-builder)
Adding a new tool (use juliaz-tool-builder)
Debugging infrastructure issues (use juliaz-debug)

Methodology: 4-Phase Improvement Cycle

Phase 1: Baseline & Failure Analysis

Before changing anything, establish what's broken and what works.

Identify the failure: What exact prompt/input causes the wrong behavior?
Trace the path: Which agent handles it? What tools does it have? What does its system prompt say?
Check logs: pm2 logs orchestrator, pm2 logs bridge, browser console for frontend
Classify the failure:
- Tool misuse: Agent calls wrong tool or doesn't call any tool → fix tool descriptions
- Prompt gap: Agent doesn't know about a capability → fix system prompt
- Missing tool: Agent literally can't do what's asked → add a tool (use juliaz-tool-builder)
- Architecture gap: No path exists between components → fix routing (use juliaz-multi-agent-optimize)

Phase 2: Targeted Fix

Apply the minimum change needed:

Tool description fix: Make descriptions crystal-clear about WHEN to use, WHEN NOT to use, and WHAT inputs to provide (see juliaz-tool-builder)
System prompt fix: Add missing capabilities, clarify boundaries, add examples
Routing fix: Add missing connections between components

Phase 3: Validation

Test the exact failure case that motivated the change:

Re-run the failing prompt → should now succeed
Run regression prompts → existing behavior unchanged
Check for hallucinated tool calls → agent shouldn't call tools for unrelated prompts
Check logs for errors → no silent failures

Phase 4: Document

Update the agent's system prompt version comment
Note the fix in thesis documentation if substantive

juliaz-Specific Failure Patterns

Symptom	Likely Cause	Fix
"I can't do that"	Missing tool or unclear prompt	Add tool or update prompt
Tool called but wrong result	Tool implementation bug	Fix the executor in `tools.ts`
Tool never called	Vague tool description	Rewrite description per `juliaz-tool-builder`
Agent calls wrong tool	Overlapping/ambiguous descriptions	Make descriptions mutually exclusive
Timeout on tool call	Service down or wrong URL	Check bridge/backend/cowork-mcp health
Generic/unhelpful response	System prompt too vague	Add specific guidance and examples

Key Files

Agent	Prompt	Tools	Executor
Orchestrator Julia	`julia/orchestrator/src/prompt.ts`	`julia/orchestrator/src/tools.ts`	`julia/orchestrator/src/tools.ts:executeTool()`
Frontend Julia	`julia/frontend/server.ts` (SYSTEM_PROMPT const)	`julia/frontend/server.ts` (chatTools object)	Inline in tool definitions
OpenClaw	`julia/openclaw/SOUL.md` + skill files	OpenClaw CLI built-in + skills	Skill scripts

More from this repository

same repository

juliaz-agent-architect

abzhaw/juliaz_agents

Architecture patterns and anti-patterns for Julia's multi-agent system. Trigger when making architectural decisions — adding agents, changing message flow, redesigning tool registries, or evaluating the system's structure. Also trigger for: 'architecture', 'system design', 'how should agents work together', 'add a new agent', 'agent patterns', or any structural question about the multi-agent system.

2026-03-080

juliaz-agent-builder

abzhaw/juliaz_agents

Scaffolding and creating new agents or sub-agents within the juliaz_agents ecosystem. Trigger whenever Raphael wants to create a new agent, add a new sub-agent, define a new agent's personality/identity, set up agent heartbeats, or modify an existing agent's SOUL/IDENTITY/TOOLS/AGENTS files. Also trigger when discussing agent conventions, the SOUL.md pattern, or how agents should be structured in this project. If someone says 'new agent', 'create an agent', 'agent personality', 'SOUL.md', or 'add a bot' — this is the skill to use.

2026-03-080

juliaz-debug

abzhaw/juliaz_agents

Cross-system diagnostics and troubleshooting for the juliaz_agents multi-agent system. Trigger when Raphael reports something is broken, not working, or behaving unexpectedly — messages not arriving, Julia not responding, bridge errors, orchestrator crashes, rate limits, silent failures, queue issues, or any 'why isn't X working' question. Also trigger for: 'debug', 'broken', 'not working', 'error', 'Julia isn't responding', 'messages stuck', 'bridge down', 'check logs', 'what went wrong', or any troubleshooting request. If something in the multi-agent system is misbehaving, this is the skill to reach for.

2026-03-080

juliaz-devops

abzhaw/juliaz_agents

Service management, process supervision, and operational tasks for the juliaz_agents multi-agent system. Trigger whenever Raphael asks to start, stop, restart, or check the status of any service — PM2, Docker, bridge, orchestrator, OpenClaw, cowork-mcp, frontend, or backend. Also trigger for: environment variable management, port conflicts, log tailing, deployment, health checks, 'is Julia running', 'start everything', 'check status', 'why is X down', pm2, docker compose, or any operational/infrastructure task. If the user mentions starting services, checking health, or managing processes — use this skill.

2026-03-080

juliaz-feature

abzhaw/juliaz_agents

Adding new features to the juliaz_agents system — new tools, API endpoints, frontend components, MCP tools, or capabilities. Trigger when Raphael wants to: add a new tool to the orchestrator or bridge, create a new API endpoint in the backend, add a frontend component or page, extend cowork-mcp with new Claude tools, add a new OpenClaw skill, or implement any new capability across the system. Also trigger for 'add a tool', 'new endpoint', 'new page', 'new component', 'extend Julia', 'Julia should be able to...', or any request to add functionality to the multi-agent system.

2026-03-080

juliaz-system

abzhaw/juliaz_agents

Foundational knowledge about the juliaz_agents multi-agent system architecture. ALWAYS trigger this skill when the conversation involves ANY component of Julia's agent ecosystem — orchestrator, bridge, cowork-mcp, OpenClaw, frontend, backend, ADHD agent, julia_medium_agent, or thesis workspace. Also trigger when Raphael asks 'where is X', 'how does X work', 'what connects to what', or any architectural/navigation question about the system. This is the map of the entire codebase and how the pieces fit together.

2026-03-080

Source

abzhaw

abzhaw/juliaz_agents

View GitHub Repository View Creator Repositories

Install command

Download

Run Skill in Manus

Useful forSOC

Software DevelopersComputer and Mathematical Occupations15-1252L4

name	juliaz-agent-improve
description	Systematic improvement of Julia's agents through performance analysis, prompt engineering, and validation. Trigger when optimizing any agent's behavior — orchestrator, frontend Julia, OpenClaw relay, or any sub-agent. Also trigger for: 'improve agent', 'agent not working well', 'optimize prompt', 'tool not being used correctly', 'Julia gives wrong answers', or any agent quality issue.

juliaz-agent-improve — Agent Performance Optimization

Systematic improvement of agents in the juliaz_agents system. Adapted from agent-orchestration-improve-agent.

Use this skill when

Improving an existing agent's performance or reliability (orchestrator, frontend Julia, OpenClaw)
Analyzing why an agent misuses tools, gives wrong answers, or fails silently
Optimizing system prompts for better tool selection
Validating agent changes before deployment

Do not use this skill when

Building a brand-new agent from scratch (use juliaz-agent-builder)
Adding a new tool (use juliaz-tool-builder)
Debugging infrastructure issues (use juliaz-debug)

Methodology: 4-Phase Improvement Cycle

Phase 1: Baseline & Failure Analysis

Before changing anything, establish what's broken and what works.

Identify the failure: What exact prompt/input causes the wrong behavior?
Trace the path: Which agent handles it? What tools does it have? What does its system prompt say?
Check logs: pm2 logs orchestrator, pm2 logs bridge, browser console for frontend
Classify the failure:
- Tool misuse: Agent calls wrong tool or doesn't call any tool → fix tool descriptions
- Prompt gap: Agent doesn't know about a capability → fix system prompt
- Missing tool: Agent literally can't do what's asked → add a tool (use juliaz-tool-builder)
- Architecture gap: No path exists between components → fix routing (use juliaz-multi-agent-optimize)

Phase 2: Targeted Fix

Apply the minimum change needed:

Tool description fix: Make descriptions crystal-clear about WHEN to use, WHEN NOT to use, and WHAT inputs to provide (see juliaz-tool-builder)
System prompt fix: Add missing capabilities, clarify boundaries, add examples
Routing fix: Add missing connections between components

Phase 3: Validation

Test the exact failure case that motivated the change:

Re-run the failing prompt → should now succeed
Run regression prompts → existing behavior unchanged
Check for hallucinated tool calls → agent shouldn't call tools for unrelated prompts
Check logs for errors → no silent failures

Phase 4: Document

Update the agent's system prompt version comment
Note the fix in thesis documentation if substantive

juliaz-Specific Failure Patterns

Symptom	Likely Cause	Fix
"I can't do that"	Missing tool or unclear prompt	Add tool or update prompt
Tool called but wrong result	Tool implementation bug	Fix the executor in `tools.ts`
Tool never called	Vague tool description	Rewrite description per `juliaz-tool-builder`
Agent calls wrong tool	Overlapping/ambiguous descriptions	Make descriptions mutually exclusive
Timeout on tool call	Service down or wrong URL	Check bridge/backend/cowork-mcp health
Generic/unhelpful response	System prompt too vague	Add specific guidance and examples

Key Files

Agent	Prompt	Tools	Executor
Orchestrator Julia	`julia/orchestrator/src/prompt.ts`	`julia/orchestrator/src/tools.ts`	`julia/orchestrator/src/tools.ts:executeTool()`
Frontend Julia	`julia/frontend/server.ts` (SYSTEM_PROMPT const)	`julia/frontend/server.ts` (chatTools object)	Inline in tool definitions
OpenClaw	`julia/openclaw/SOUL.md` + skill files	OpenClaw CLI built-in + skills	Skill scripts