agentleFS
mikezamayiasmikezamayias/agent-skills/skills/answer-sheet/SKILL.md

answer-sheet

Skill1 starsChanged 26 days agoBSD-3-ClauseView on GitHub

Puts open questions for the user on one self-contained HTML page, with answers saved to a dated Markdown file. Use when more than a handful of questions, decisions or sign-offs are waiting instead of a coded list in chat. Also use when the decision-maker cannot keep up with codes, or a review round needs screenshots per question.

The file

Show its rules and text

We split this skill into the rules it gives an agent

21 rules. About 0 are common, 11 are distinctive, and 10 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈11

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 10

    Very specific

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

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

    Write <questions-dir>/<YYYY-MM-DD>.json in the format of scripts/example.json.

    line 36
  2. Very specific:

    Add a suffix (-b, -c) for a second sheet on the same day.

    line 37
  3. Very specific:

    Start the server once per session in the background with python3 <skill>/scripts/answer_sheet.py serve <questions-dir> <answers-dir> [port] …

    line 39
  4. Very specific:

    If the Mac runs Tailscale Serve, check tailscale serve status first.

    line 41
  5. Very specific:

    When it already proxies an HTTPS port to 127.0.0.1:8765, keep the default bind, because Tailscale holds that port on the Tailscale address …

    line 42
  6. Very specific:

    Always give the user the Tailscale URL, such as https://<machine>.<tailnet>.ts.net:8765/<date>.html, and check it returns 200 before …

    line 44
  7. Very specific:

    Use tailscale status --self for the machine name, because an old -1 name can still appear in the serve list.

    line 45
  8. Distinctive:

    Do not repeat the questions in chat.

    line 48
  9. Very specific:

    When the user says they saved, read <answers-dir>/<date>.md, record each ruling where the project keeps decisions (see the decision-sweep …

    line 49
  10. Distinctive:

    If the answers raise new questions, they go on the next sheet.

    line 50
  11. Distinctive:

    Start from scratch: what exists today, what the problem is, and who it affects.

    line 59
  12. Distinctive:

    When asking for a sign-off, quote the exact sentence or show exactly what changed.

    line 62
  13. Distinctive:

    Use JPEG. Some model APIs reject large PNG exports, and JPEG keeps the repository small.

    line 64
  14. Distinctive:

    Every question also gets a free-text box, so there is no need for an "Other" option.

    line 67
  15. Distinctive:

    Never ask the decision-maker to check a fact you can check yourself.

    line 68
  16. Distinctive:

    Check it, then ask only the decision.

    line 69
  17. Distinctive:

    Never ask about a concern the decision-maker has already dismissed.

    line 70
  18. Distinctive:

    If the server is down, "Copy answers" puts the same Markdown on the clipboard for pasting into chat.

    line 79
  19. Very specific:

    Check with curl -s -o /dev/null -w '%{http_code}' http://127.0.0.1:8765/ before starting one.

    line 85
  20. Very specific:

    Only bold and bullets that start with a hyphen are formatted.

    line 87
  21. Distinctive:

    Keep emoji out of sheets.

    line 88

≈ “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

Answer sheet

A question in chat competes with everything else in the chat. Long lists of coded questions become unreadable once the reader forgets what each code meant. An answer sheet replaces the chat list with one page per round: every question carries its own background, the options and what each leads to, a recommendation, and a free-text box. The page saves answers in the browser as the user goes and writes them to the repository on "Save answers".

When to use it

  • More than about five open questions, or any sentence-level sign-off.
  • The questions span sessions, so the reader will not remember the earlier messages.
  • A question needs a picture, such as a design frame, a chart, or a screenshot.

For one or two quick questions, ask in chat, with each code followed by its meaning.

Files

  • scripts/answer_sheet.py builds and serves the page. It uses only the Python standard library.
  • scripts/example.json is a template question file.
  • scripts/test_answer_sheet.py is the check. Run it after changing the script.

Pick two directories in the project, for example questions/ and answers/, and record them in the project's AGENTS.md so every agent uses the same place. The answers directory is the decision-maker's. Agents read it and never edit it.

Steps

  1. Write <questions-dir>/<YYYY-MM-DD>.json in the format of scripts/example.json. Add a suffix (-b, -c) for a second sheet on the same day.
  2. Build the page with python3 <skill>/scripts/answer_sheet.py build <questions-dir>/<date>.json.
  3. Start the server once per session in the background with python3 <skill>/scripts/answer_sheet.py serve <questions-dir> <answers-dir> [port] [bind]. It binds to 127.0.0.1 on port 8765 by default. If the Mac runs Tailscale Serve, check tailscale serve status first. When it already proxies an HTTPS port to 127.0.0.1:8765, keep the default bind, because Tailscale holds that port on the Tailscale address and a second bind there fails. Otherwise pass the Mac's Tailscale address as the bind argument, so Save still works from a phone.
  4. Always give the user the Tailscale URL, such as https://<machine>.<tailnet>.ts.net:8765/<date>.html, and check it returns 200 before sending it. Use tailscale status --self for the machine name, because an old -1 name can still appear in the serve list. Also open http://127.0.0.1:8765/<date>.html in the in-app preview. In chat, say in one or two sentences what the sheet covers. Do not repeat the questions in chat.
  5. When the user says they saved, read <answers-dir>/<date>.md, record each ruling where the project keeps decisions (see the decision-sweep skill), act on it, and commit the answer file with the changes.
  6. If the answers raise new questions, they go on the next sheet.

Writing a question

Each question must stand on its own for a reader who has forgotten everything said before.

  • Title: the decision in plain words, as a question or a short instruction. Never a code.
  • Context: paragraphs and bullets that start with a hyphen, with **bold** allowed.
    • Start from scratch: what exists today, what the problem is, and who it affects.
    • Explain every product name, acronym, and legal term the first time it appears.
    • State the harm the recommendation prevents, not only what it does.
    • When asking for a sign-off, quote the exact sentence or show exactly what changed.
  • Images: optional, with src relative to the questions directory and a caption. Use JPEG. Some model APIs reject large PNG exports, and JPEG keeps the repository small.
  • Options: 2 to 4, each with a label and a detail saying what happens next if it is chosen. Mark exactly one "recommended": true, unless you genuinely have no view. Every question also gets a free-text box, so there is no need for an "Other" option.
  • Never ask the decision-maker to check a fact you can check yourself. Check it, then ask only the decision.
  • Never ask about a concern the decision-maker has already dismissed.

Behaviour to rely on

  • Answers autosave in the browser, keyed by date and title, so a reload loses nothing.
  • Save writes a block that starts ## <title> (answered in the answer sheet). A second Save of the same sheet replaces that block instead of appending a duplicate.
  • The file may hold the user's own notes too. The server only touches its own block.
  • If the server is down, "Copy answers" puts the same Markdown on the clipboard for pasting into chat.
  • window.answerSheet() returns the current state, which is useful when driving the page from browser automation.

Pitfalls

  • Starting a second server on the same port fails. Check with curl -s -o /dev/null -w '%{http_code}' http://127.0.0.1:8765/ before starting one.
  • The rendered page escapes HTML. Only **bold** and bullets that start with a hyphen are formatted.
  • Keep emoji out of sheets.

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.