CtrlK
BlogDocsLog inGet started
Tessl Logo

agent-md-refactor

Refactor bloated AGENTS.md, CLAUDE.md, or similar agent instruction files to follow progressive disclosure principles. Splits monolithic files into organized, linked documentation.

81

1.36x
Quality

72%

Does it follow best practices?

Impact

94%

1.36x

Average score across 3 eval scenarios

SecuritybySnyk

Passed

No known issues

Optimize this skill with Tessl

npx tessl skill review --optimize ./dist/plugins/agent-md-refactor/skills/agent-md-refactor/SKILL.md
SKILL.md
Quality
Evals
Security

Evaluation results

96%

40%

Clean Up a Messy Agent Config File

Contradiction detection and root minimization

Criteria
Without context
With context

Both contradictions found

100%

100%

Contradiction format: Instruction A/B

30%

100%

Contradiction has resolution question

30%

100%

Root file under 50 lines

0%

100%

Project description in root

100%

100%

Non-standard package manager in root

100%

100%

Non-standard commands in root

100%

100%

Non-essential content moved out

40%

100%

Linked files created

100%

100%

Root has Quick Reference and Detailed Instructions sections

0%

50%

All phases addressed

50%

100%

87%

34%

Reorganize Sprawling Agent Instructions for a Full-Stack Project

File structure and linked file templates

Criteria
Without context
With context

Subdirectory location

30%

100%

3–8 linked files

0%

100%

Topic-named files

100%

100%

Linked file has Overview section

0%

100%

Linked file has Rules section

0%

70%

Linked file has Examples section

0%

0%

Root has Detailed Instructions links

100%

100%

No deep nesting

100%

100%

Files are self-contained by topic

100%

100%

Root file minimal

100%

100%

100%

Audit and Prune a Legacy Agent Configuration File

Deletion flagging and verification

Criteria
Without context
With context

Flagged for Deletion section

100%

100%

Deletion table format

100%

100%

Vague instructions flagged

100%

100%

Redundant/obvious instructions flagged

100%

100%

Deletion reason provided

100%

100%

Specific instructions preserved

100%

100%

Root file under 50 lines

100%

100%

Links verified

100%

100%

No instruction loss confirmed

100%

100%

Linked files created

100%

100%

Repository
softaworks/agent-toolkit
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.