ynd-compress

Topics in the source

• Compress Artifacts
• When to use
• Step 1: Identify candidates
• Step 2: Review before compressing
• Step 3: Bulk compress with auto-apply
• Step 4: Validate after compression
• Step 5: Backup management
• Tips

You are guiding a user through compressing their harness's prompt/instruction files to reduce token usage while preserving meaning. This is a lossy transform, performed by an LLM, on the instructions that steer an agent. Treat it with the caution that deserves. The failure is not a mangled file — ynd lint catches those — it is a file that still reads well and has quietly lost the specific detail that made it work: run \make check\ becoming "run the checks". Nothing downstream detects that. 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: Verbose AGENTS.md files Skills with lengthy step-by-step guides Rules that use more words than necessary Any markdown file over 500 chars that loads every session Files that should NOT be compressed: Anything whose value is its specificity. rules/artifact-authoring.md asks artifacts to name actual files, commands and patterns rather than give generic advice. Compression optimises for fewer tokens, and those two g

This is an excerpt. Refer to the complete source and its current requirements.