vs-tldr
User-invoked compression of the explanation the agent just gave. Use when the user types `/vs-tldr`, says `tldr`, `too long`, `simpler`, `in plain english`, `what do you mean?`, `wait, what?`, or `lost me`; re-pitch the relevant context shorter and simpler without dropping the premise, the decision, or the caveats.
- License 许可
- MIT
- 所属仓库
- vltansky/vs
- 最近更新
- 2026-11-06
- Stars
- ★ 0
- Forks
- 1
01 前言 Front Matter
- name
- vs-tldr
- description
- User-invoked compression of the explanation the agent just gave. Use when the user types `/vs-tldr`, says `tldr`, `too long`, `simpler`, `in plain english`, `what do you mean?`, `wait, what?`, or `lost me`; re-pitch the relevant context shorter and simpler without dropping the premise, the decision, or the caveats.
02 资源文件 3
- 参考文档agents/openai.yaml169 B
- 许可证LICENSE.mattpocock-skills1.0 KB
- 脚本test/tldr.static.eval.ts2.0 KB
03 SKILL.md 正文
TLDR
Compress and simplify the explanation the user just read.
Flow Contract
- Kind: Building block
- Inputs: The current conversation and the explanation that was too long or too complex
- Outputs: A shorter, simpler re-pitch that keeps the missing premise and the load-bearing detail
- Status:
RE_PITCHEDorBLOCKED_AMBIGUOUS_TARGET - Consumers: Any active workflow or direct user invocation
- Skip conditions: Skip when the user wants a new answer, a whole-task recap, or a prose rewrite rather than compression of the current explanation
Repair the message
- Treat
/vs-tldras a length-and-complexity signal: the content was right, the delivery was too long or too dense. - Re-pitch that: look back far enough to recover the missing premise, not just the last paragraph.
- Cut hard. Drop process narration, hedging, restated context, and any option you already rejected.
- Use plain English and short sentences. Reuse terms from the applicable
CLAUDE.md,CONTEXT.md, current code, and conversation when they exist; do not invent replacement jargon. - Keep the decision, caveats, technical accuracy, and action that matter. Do not start a new task, change code, or silently remove a condition. Shorter is the goal; wrong is not the price.
- Return the repair directly, without a preamble, process narration, or duplicate recap.
- If the target is genuinely ambiguous, ask one compact question. Otherwise infer it from the current turn and continue.
Check the result
- It is materially shorter than what it replaces.
- It is simpler, not just blunter — a reader who missed the premise now has it.
- Project nouns replace invented ones.
- No caveat, condition, or action was lost in the cut.
- A second
/vs-tldrcompresses again rather than degrading into fragments. - After the re-pitch, if a from-zero visual explainer would help, propose
/vs-eli5in one short line. Do not run it.
Boundary
Use /vs-tldr at any point, including inside another skill. It compresses one
explanation after the fact; it does not replace /vs-recap for the whole current
situation, /vs-write for editing supplied prose, or /vs-eli5 for a from-zero visual explainer.
This skill stays prose-only. It does not become HTMDX.
Apply the shared output style to the re-pitch. This skill is intentionally small: the leading signal is that this specific explanation was too long or too complex, not a general request to make every response shorter.
Workflow
Direct: emit Next only. Composed: return to caller.
Prev: explanation that was too long or too complex | any active skill Next: done Relevant: none
04 同仓库更多 Skills
vs-write
Use when asked to write, rewrite, edit, tighten, simplify, or polish prose for clarity, or to put text in STE, ASD-STE100, or controlled language. Produces plain, direct, accessible copy without losing substance or precision.
vs-fake-verify
暂无简介
vs-fake-verify
暂无简介
vs-verify
Use when asked to verify, prove it works, check completion, or before claiming done. Produces evidence-backed PASS/WARN/FAIL status.
vs-fake-verify
暂无简介
vs-verify
Use when asked to verify, prove it works, check completion, or before claiming done. Produces evidence-backed PASS/WARN/FAIL status.