dontbmad-compress-artifacts
Compress BMAD planning artifacts (PRDs, architecture docs, stories, project-context) into terse format to reduce input tokens when agents read them. Originals backed up as .original.md. Use when user says "compress artifacts", "compress planning docs", or "shrink the docs".
- License 许可
- MIT
- 最近更新
- 2026-06-05
- Stars
- ★ 0
- Forks
- 1
01 前言 Front Matter
- name
- dontbmad-compress-artifacts
- description
- Compress BMAD planning artifacts (PRDs, architecture docs, stories, project-context) into terse format to reduce input tokens when agents read them. Originals backed up as .original.md. Use when user says "compress artifacts", "compress planning docs", or "shrink the docs".
02 资源文件 1
- 参考文档bmad-skill-manifest.yaml327 B
03 SKILL.md 正文
Compress BMAD Artifacts
Compress planning artifact markdown files into terse caveman format so agents load fewer input tokens in future sessions. Human-readable originals are backed up as <filename>.original.md.
When to Use
- After generating multiple planning artifacts (PRDs, architecture, stories, epics)
- Before a heavy multi-agent session (party mode, dev-story) to reduce context load
- When context window pressure is high and you need to reclaim tokens
What Gets Compressed
Scan {planning_artifacts}/ (resolved from {project-root}/_bmad/bmm/config.yaml) for .md files. Also check for **/project-context.md.
Compress: Prose sections in markdown files — descriptions, rationale, notes, acceptance criteria written in sentences.
Never compress:
- Files already ending in
.original.md(skip — these are backups) - Code blocks (fenced ``` and indented)
- Inline code (
backtick content) - URLs, file paths, commands
- Technical terms, proper nouns, version numbers
- YAML frontmatter
- Table structure (compress cell text, keep structure)
- Mermaid diagrams
Process
For each .md file found:
- Check if
<filename>.original.mdalready exists. If yes, skip (already compressed). - Copy the current file to
<filename>.original.mdas human-readable backup. - Rewrite the original file in place using caveman compression rules:
- Remove articles (a/an/the), filler, hedging, pleasantries
- Use fragments and short synonyms
- Merge redundant bullets that say the same thing differently
- Keep one example where multiple examples show the same pattern
- Preserve all markdown structure (headings, bullets, numbering, tables)
- Preserve all code blocks and inline code exactly
- Report: filename, original token count (approx), compressed token count, savings percentage.
Compression Rules
Remove
- Articles: a, an, the
- Filler: just, really, basically, actually, simply, essentially, generally
- Pleasantries: "sure", "certainly", "of course"
- Hedging: "it might be worth", "you could consider", "it would be good to"
- Redundant phrasing: "in order to" -> "to", "make sure to" -> "ensure", "the reason is because" -> "because"
- Connective fluff: "however", "furthermore", "additionally"
Preserve Exactly
- Code blocks (fenced and indented)
- Inline code
- URLs and markdown links
- File paths and commands
- Technical terms, library names, API names
- Proper nouns, dates, version numbers
- Environment variables
Compress
- Short synonyms: "big" not "extensive", "fix" not "implement a solution for", "use" not "utilize"
- Fragments OK: "Run tests before commit" not "You should always run tests before committing"
- Drop "you should", "make sure to", "remember to" — state the action directly
Example
Before:
The application should provide users with the ability to search for available delivery slots in their area. The system must validate that the selected time slot is still available before confirming the booking.
After:
Users search available delivery slots by area. System validate slot availability before booking confirmation.
Restore
To restore a file to its original version:
mv <filename>.original.md <filename>.md
Output
After compression, print a summary table:
| File | Before | After | Saved |
|---|---|---|---|
| prd.md | ~2400 tokens | ~1300 tokens | 46% |
| architecture.md | ~3100 tokens | ~1700 tokens | 45% |
| Total | ~5500 | ~3000 | ~45% |
04 同仓库更多 Skills
dontbmad-grill
Interview the user relentlessly on a plan, decision, or artifact until every branch of the decision tree is resolved. One question at a time, with a recommended answer for each. Reads existing artifacts and codebase before asking. Use when the user says "grill me", "stress-test this plan", "interrogate this", or when invoked from a BMAD planning skill (architecture, PRD, story, UX) before finalizing the artifact.
dontbmad-graphify
暂无简介
dontbmad-graphify-query
暂无简介
dontbmad-graphify-path
暂无简介
dontbmad-graphify-explain
暂无简介
dontbmad-claude-design-to-bmad-uiux
Convert Claude Design output into a BMAD-compatible UI/UX handoff specification. Use when the user wants Claude Design to BMAD UX export, AC traceability, or ready-for-dev UIUX format.