agentleFS
leolou0512leolou0512/claude-skills/skills/qu/SKILL.md

qu

Skill1 starsChanged 26 days agoMITView on GitHub

Quick update — the ultra-terse status format for when the user comes back after the agent has been working alone for a while. Three fixed headers only: RUNNING / RESULTS / NEED YOU. Use when the user says "qu", "/qu", "quick update", "quick status", "what's up", "back, anything?", or returns from an absence and asks how it went. Also the default register for any unprompted update after a long autonomous stretch.

The file

We split this skill into the rules it gives an agent

2 rules. About 0 are common, 2 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈2

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Every section present, even if the content is "(nothing)".

    line 41
  2. Distinctive:

    If they ask a follow-up, answer that one thing in ≤3 lines — do not re-emit the block.

    line 51

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

/qu — quick update

The user has been away; they want the state in five seconds. The failure mode is a block of prose. Caveman register (fragments OK, no filler, no hedging), three headers, done.

Before writing (≤1 min, no audit)

Refresh from ground truth, not memory: nvidia-smi, pgrep/ps, the run's state.txt/log tail, output files. If a real number takes >1 min to pull, write "not pulled yet" — never estimate as fact. Heavy digging → subagent, you format.

Format — exactly this, nothing before or after

RUNNING
- <job/run, plain name> — <state>, <progress> — ETA <HH:MMZ or "unknown">
- (nothing) / DEAD: <what>, <cause> — <salvaged / relaunched?>

RESULTS
- <headline number first, exact, with the baseline it beats/misses>
- <link or path to the file>   ← only if a file exists
- (none yet — <why, 5 words>)

NEED YOU
- <the one decision/blocker + who owns it> / (nothing)

Hard rules

  • ≤ 10 lines total. One line per item. No paragraphs. No preamble, no sign-off, no "summary" sentence, no headers beyond the three.
  • Every section present, even if the content is "(nothing)". Empty ≠ omitted.
  • Dead/failed jobs go in RUNNING as DEAD: — first bullet of that section. Never buried, never softened.
  • Numbers exact, in %, with the comparison inline ("12.6% mSUN vs 12.56 stock").
  • ETAs absolute clock + Z. Never "soon"/"in a bit".
  • Plain names: a run codename gets a 2–4 word gloss in parentheses on first use ("zeroinit_final (epoch-656 chain)"). No session shorthand, no phase labels.
  • NEED YOU holds at most the top 1–2 items. If the user needs to choose, state the default you'll take if they say nothing.
  • Argument narrows scope: "qu grid" → grid only, same format.
  • If they ask a follow-up, answer that one thing in ≤3 lines — do not re-emit the block.

Example

RUNNING
- coord_1p2_lat_1p5_iso (isotropic-noise diffusion) — epoch 412/1000, loss 5.31 — ETA 08-16 03:40Z
- DEAD: eval on node A — OOM at relax stage 04:12Z — relaunched batch 8, ETA 06:30Z

RESULTS
- epoch-656 chain: mSUN 13.1% (327/2500) vs 12.56% stock baseline — raw, pre-relax
- reports/2026-08-15_zeroinit_e656.md

NEED YOU
- Relax numbers land ~06:30Z; if <12.56% I stop the chain unless you say otherwise.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.