From ynh
Workflow for compressing prompt and instruction files using LLM-powered techniques, with backup management and restore.
How this skill is triggered — by the user, by Claude, or both
Slash command
/ynh:ynd-compressThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
You are guiding a user through compressing their harness's prompt/instruction files to reduce token usage while preserving meaning.
You are guiding a user through compressing their harness's prompt/instruction files to reduce token usage while preserving meaning.
Use after authoring or updating skills, agents, rules, or instructions. Compression reduces token count for files that will be loaded into every AI session. Particularly valuable for verbose instructions or detailed skills.
Help the user find files worth compressing. Good candidates are:
AGENTS.md filesFiles that should NOT be compressed:
Start with interactive mode so they can review the compression:
ynd compress skills/code-review/SKILL.md
This shows the original and compressed versions side by side with the reduction percentage. They can accept or skip.
Once they trust the quality, compress multiple files at once:
ynd compress -y skills/*/SKILL.md agents/*.md rules/*.md
Or compress everything discovered automatically:
ynd compress -y
Compression preserves YAML frontmatter structurally (it's never sent to the LLM), but always validate after:
ynd validate
ynd lint
Both should pass. If they don't, restore from backup and report the issue.
Every compression creates an automatic backup in ~/.ynd/backups/. Show the user how to manage them:
# See what backups exist for a file
ynd compress --list-backups skills/code-review/SKILL.md
# Restore the most recent backup
ynd compress --restore skills/code-review/SKILL.md
# Restore a specific older version
ynd compress --list-backups skills/code-review/SKILL.md # note the number
ynd compress --restore --pick 3 skills/code-review/SKILL.md
ynd compress -v claude to force a specific vendor if auto-detect picks the wrong one.npx claudepluginhub eyelock/ynhProvides behavioral guidelines to reduce common LLM coding mistakes, focusing on simplicity, surgical changes, assumption surfacing, and verifiable success criteria.
Searches, retrieves, and installs Agent Skills from prompts.chat registry using MCP tools like search_skills and get_skill. Activates for finding skills, browsing catalogs, or extending Claude.
Creates, edits, and optimizes skills for Claude Code, including drafting, evaluating with test prompts, iterating on performance, and improving skill descriptions for better triggering accuracy.