CtrlK
BlogDocsLog inGet started
Tessl Logo

writing-agents

Use when creating new agents, editing existing agents, or defining specialized subagent roles for the Task tool

Install with Tessl CLI

npx tessl i github:aaddrick/claude-desktop-debian --skill writing-agents
What are skills?

78

1.14x

Quality

67%

Does it follow best practices?

Impact

97%

1.14x

Average score across 3 eval scenarios

Optimize this skill with Tessl

npx tessl skill review --optimize ./.claude/skills/writing-agents/SKILL.md
SKILL.md
Review
Evals

Evaluation results

100%

20%

Build Automation Specialist Agent

Agent structure and persona quality

Criteria
Without context
With context

Correct file location

100%

100%

YAML frontmatter present

100%

100%

Specific persona

100%

100%

Project-specific persona content

100%

100%

Explicit scope boundaries

0%

100%

Anti-patterns section present

66%

100%

Specific anti-patterns

100%

100%

Project-derived anti-patterns

100%

100%

Project structure included

100%

100%

Session restart notice

0%

100%

SETUP.md created

100%

100%

Description has trigger conditions

100%

100%

Without context: $0.3178 · 1m 32s · 17 turns · 22 in / 5,047 out tokens

With context: $0.5472 · 1m 52s · 24 turns · 457 in / 6,020 out tokens

100%

Agent Suite for a Full-Stack Development Team

Description format and model selection

Criteria
Without context
With context

All three agents created

100%

100%

Description includes triggers

100%

100%

No process in formatter description

100%

100%

No process in architect description

100%

100%

No process in reviewer description

100%

100%

Formatter uses haiku

100%

100%

Architect uses opus

100%

100%

Descriptions under 500 chars

100%

100%

YAML frontmatter on all agents

100%

100%

Description role statement

100%

100%

Correct file placement

100%

100%

Without context: $0.3627 · 1m 36s · 23 turns · 27 in / 4,973 out tokens

With context: $0.5630 · 1m 58s · 25 turns · 458 in / 6,312 out tokens

92%

17%

Orchestrator Agent for Feature Delivery Pipeline

Multi-agent coordination protocols

Criteria
Without context
With context

Orchestrator file created

100%

100%

Orchestrator does not write code

100%

100%

Delegates to frontend-specialist

100%

100%

Delegates to backend-specialist

100%

100%

Delegation format defined

100%

100%

Report format specified

50%

100%

Orchestrator scope boundaries

100%

100%

No deploy without approval

0%

0%

coordination-guide.md created

100%

100%

YAML frontmatter on orchestrator

100%

100%

Orchestrator model is opus

0%

100%

Without context: $0.2579 · 1m 24s · 13 turns · 16 in / 4,356 out tokens

With context: $0.5116 · 1m 59s · 21 turns · 208 in / 5,857 out tokens

Evaluated
Agent
Claude Code
Model
Claude Sonnet 4.6

Table of Contents

Is this your skill?

If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.