agentleFS
robin-blocksrobin-blocks/otra-city/.claude/skills/asks/SKILL.md

asks

Skill0 starsChanged 2 months agoMITView on GitHub

Plain-language status for Robin — where the session got to, and every task that is waiting on him, each one summarised simply with step-by-step instructions he can follow without reading the code. Use whenever Robin types /asks or asks any version of "what's on me?", "what do I need to do?", "what am I blocking?", "where are we?", "what's left for me?", "anything waiting on me?", "catch me up". Unlike /tidy, this changes nothing and closes nothing down — it is a briefing, and it can be run at any point in a session, including the middle.

1 other project uses a version of this.

Adopted by

Most recently taken up by

  • Other versions1 project uses a version of this.

    Word for word · 0

    None found.

    Edited · 0

    None found.

    Wrote their own version · 1

    1. EdwardAF-IT/claude-skills★ 1same name

    How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.

  • In this repository2 other files it gives its agents.

The file

We split this skill into the rules it gives an agent

7 rules. About 0 are common, 6 are distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈6

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 1

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Read-only — it changes nothing, writes nothing, and never runs a build.

    line 31
  2. Distinctive:

    Do this before writing anything, because it is the check most likely to change the report, and the one nothing else catches.

    line 40
  3. Distinctive:

    If it does not appear, check the file itself:

    line 59
  4. Distinctive:

    Leave these out. Padding the list is how the real items get skimmed past.

    line 88
  5. Distinctive:

    Never invent a task: If the list is empty, say "nothing is waiting on you" and stop.

    line 155
  6. Distinctive:

    Check before you claim: If you say a thing is live, you have loaded it.

    line 161
  7. Distinctive:

    If you say a fix landed, you have looked inside main for it.

    line 162

≈ “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

/asks — what's waiting on Robin, in plain words

The point

Robin has been away, or deep in something else, or has just come back to a session that has been running for hours. He needs two things and nothing else:

  1. Where are we? One short paragraph. Not a changelog.
  2. What is waiting on me, and exactly how do I do each one?

The failure this prevents is not "a task was forgotten". It is a task that only Robin can do, sitting unnoticed while everyone assumes it is done. A decision nobody chased. A deploy that never happened, so the fix is written but not live. A draft addressed to somebody outside the project that was never sent, so they are still waiting for a reply.

Those are invisible from the inside. The code looks finished. The PR is merged. Nothing is red. And the thing still is not done.

Run the state script first

bash .claude/skills/asks/scripts/state.sh

Read-only — it changes nothing, writes nothing, and never runs a build. It gathers the mechanical half: branch and working tree, open PRs and their checks, [ROBIN] markers in docs/, unsent drafts, whether the live site answers, and which local servers are running.

It gathers facts. It does not know what matters. That part is yours.

Then check that what looks done actually landed

Do this before writing anything, because it is the check most likely to change the report, and the one nothing else catches.

For each thing this session believes it finished, confirm it is really on main and really live — do not trust a merged PR:

git fetch origin -q && git log --oneline origin/main -8

Then look inside main for the actual change, not just the commit title. A squash merge captures the branch as it was at the moment the button was pressed; a commit pushed a minute later is silently left behind, and the PR still says "merged".

git log origin/main --oneline --grep="<a distinctive word from the commit>"

If it does not appear, check the file itself:

git show origin/main:<path/to/file> | grep -c "<a symbol the change added>"

A zero there means the work is orphaned on a branch. That is a task for Robin, and a high-priority one, because from the outside it looks complete.

Where the tasks on Robin come from

Five places. Sweep all five; most sessions have tasks in two or three.

  • The conversation. Things he was asked to decide and has not answered. Offers made that he never took up. Corrections he half-gave. Read the whole session, not the last few messages.
  • [ROBIN] markers in docs/. This repo marks maintainer decisions in place. The script lists them. Read the surrounding paragraph for each — the marker alone is not enough context to act on.
  • Things only he has access to. Deploys, DNS, accounts, credentials, paid services, anything on a machine or dashboard that is his. If a fix is written but needs a deploy he has to run, the task is the deploy, not the fix.
  • Things only he can say. Replies to people outside the project, business terms, credit and naming, whether to take a piece of work on at all. Taste calls: does this look right.
  • Work left orphaned, per the section above.

What is not a task on Robin

Leave these out. Padding the list is how the real items get skimmed past.

  • Anything the assistant can do itself and simply has not yet. Say it is in progress under "where we are", or just do it.
  • Anything already done. A merged PR is not a task.
  • Suggestions and nice-to-haves he has not asked for. If it is worth raising, raise it as one line at the end under "worth knowing", not as a task.
  • Anything waiting on a third party. That is their task; note it as blocked.

Write it in plain language

This is the part that makes the skill worth having, so hold the line on it. Robin knows this codebase, but he is reading this cold, possibly on a phone, possibly to decide whether he has ten minutes to spare.

  • Say what a thing is before you name it. Not "the observe flag" but "the setting that stops the broadcast camera appearing to visitors as a person standing on the pitch (observe)".
  • No unexplained jargon. Not "squash merge dropped the commit" but "the merge only picked up the changes as they were a minute earlier, so the last fix got left behind".
  • Numbers only when they change what he does. "20× inside budget" earns its place; "p95 4.3 ms" does not.
  • Say why it matters, in one line, in terms of consequence. What actually goes wrong if he never does it. Not "this is important".
  • Say how long it will take. Rough is fine: seconds, a few minutes, needs a sitting-down. He is triaging.
  • Never use section numbers from a document as though they mean something. Not "§5.2 is unresolved" but "whether to freeze the stadium for a month at a time".

The shape of the report

Two parts. Keep the first short — the second is the point.

Part 1 — Where we are

One paragraph, four or five sentences at most. What got done, what is on main and live, what is still moving. No lists. If the session is mid-flight and something is running, say so and say what happens when it finishes.

Part 2 — What's on you

Ordered by consequence, worst first — not by how long they take.

For each task:

  • A short title that names the actual thing, readable on its own.
  • One or two lines of context. What this is and why it is waiting on him specifically. Include what breaks or stays broken if it never happens.
  • Numbered steps. Each step is one action, small enough to do without thinking about it. Where a step is a command, put it in its own fenced bash block, one command per block — the app puts a Run button on those, so a block with two commands in it costs him a copy-paste.
  • What to expect. How he knows it worked, and what to do if it did not.
  • Roughly how long.

Where a task is a decision rather than an action, do not fake steps. Give him: what the choice is, the options in plain terms, what each one costs him later, which one you would pick and why, and where to record the answer once he has made it.

End with "Worth knowing" — at most three one-line notes about things that are true but not tasks. Then stop.

Rules

  • Never invent a task. If the list is empty, say "nothing is waiting on you" and stop. An honest empty list is the most useful possible output.
  • Never turn one task into three by splitting a single action into sub-tasks. He counts items to decide whether to start.
  • Do not re-explain work he has already been told about. Link to the file or the PR and move on.
  • Check before you claim. If you say a thing is live, you have loaded it. If you say a fix landed, you have looked inside main for it. Everything in this report is checkable, so check it.
  • Say plainly when you are unsure which of two things he meant, rather than picking one and presenting it as settled.

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.