writing-skills

Use when creating new skills, editing existing skills, or verifying skills work before deployment

Install with Tessl CLI

npx tessl i github:aaddrick/claude-desktop-debian --skill writing-skills

What are skills?

1.29x

Quality

58%

Does it follow best practices?

Impact

100%

1.29x

Average score across 3 eval scenarios

Optimize this skill with Tessl

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

SKILL.md

Review

Evals

Evaluation results

100%

39%

Skill: Async Retry with Exponential Backoff

Skill YAML frontmatter and CSO description rules

Criteria

Without context

With context

Description starts with 'Use when'

100%

Description is triggers-only

33%

100%

Description in third person

100%

Description under 500 characters

100%

Name format valid

100%

Only two YAML fields

30%

100%

YAML under 1024 chars

100%

Keywords for discoverability

100%

Single language code example

100%

No narrative storytelling

100%

Without context: $0.7932 · 2m 52s · 32 turns · 240 in / 8,040 out tokens

With context: $0.3631 · 1m 22s · 17 turns · 302 in / 4,071 out tokens

100%

Skill: CSV Processing with Python's csv Module

File organization and token efficiency

Criteria

Without context

With context

Heavy reference in separate file

100%

Short patterns inline

100%

Flowchart for decision only

100%

No generic flowchart labels

100%

Flowchart not used for reference

100%

SKILL.md word count

100%

Single language examples

100%

No fill-in-the-blank templates

100%

Separate file linked from SKILL.md

100%

Reference file over 100 lines

100%

Without context: $0.6805 · 3m 39s · 22 turns · 1,113 in / 12,183 out tokens

With context: $0.8296 · 3m 7s · 29 turns · 65 in / 10,950 out tokens

100%

28%

Skill: Always Verify Before Closing a GitHub Issue

Discipline skill rationalization resistance

Criteria

Without context

With context

Rationalization table present

66%

100%

Red flags list present

100%

Explicit loophole closures

75%

100%

Spirit vs letter addressed

100%

3+ simultaneous pressures per scenario

100%

Concrete A/B/C options

100%

Real constraints in scenarios

87%

100%

Core rule stated clearly

100%

No narrative storytelling

100%

Two pressure scenarios present

100%

Without context: $0.3231 · 2m 6s · 15 turns · 20 in / 5,402 out tokens

With context: $0.6492 · 2m 20s · 23 turns · 2,634 in / 6,515 out tokens

Evaluated: 4 days ago
Agent: Claude Code
Model: Claude Sonnet 4.6

Table of Contents

Skill: Async Retry with Exponential Backoff Skill: CSV Processing with Python's csv Module Skill: Always Verify Before Closing a GitHub Issue

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.