agentleFS
KrishP147KrishP147/skills/grill-docs/SKILL.md

grill-docs

Skill37 starsChanged 47 days agoMITView on GitHub

Grill the user about a plan (same interview as grilling) and record it — Q→A transcript in grills/, each decision appended to the repo's decisions doc (both under skilleddocs/ by default). Use for "grill me and document it", "grill with docs".

  • Installs packages

2 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

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

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈1

    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:

    Call the Skill tool with "grilling" and run that interview unchanged.

    line 7
  2. Distinctive:

    Every answer is a decision, delegated ones too: planners, verifiers and consult-plan read the register.

    line 58

≈ “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

Call the Skill tool with "grilling" and run that interview unchanged. Your addition: write down each answer as it comes in, mid-round too, so an abandoned session still leaves a record.

Paths (skilleddocs/ is at the repo root, per TRACKING.md):

  • transcript: skilleddocs/grills/<local-date>-<slug>.md. Slug: the argument's last word only when it is kebab-case and follows a doc path or a quoted subject (docs/plan.md auth-plan, "our caching plan" cache-plan); else 2–4 topic words, and the whole argument is the subject (should we support e-books keeps e-books).
  • register (the decisions doc): the one the repo's AGENTS.md/CLAUDE.md names (several name one, e.g. a monorepo: the nearest walking up from the working dir to the repo root wins, and say which others you skipped; a path it names is relative to that file's folder); else an existing docs/decisions.md; else an ADR folder (docs/adr/, doc/adr/, docs/decisions/); else skilleddocs/decisions.md. Same rule as consult-plan. An ADR folder gets one new ADR per decision in its own format, carrying the same Source; one that reverses or narrows an old ADR also sets the old one's Status to Superseded by NNNN / Amended by NNNN, the only edit allowed to an old record. Always name the chosen path in the header's Register: field.
  • register outside skilleddocs/ while skilleddocs/ stays untracked (below) → the transcript goes in grills/ beside the register file or ADR folder instead (docs/grills/… for docs/decisions.md or docs/adr/), so its Sources ship with it.

Untracked → exclude skilleddocs/ per TRACKING.md (it also covers the pip install . hazard).

Before round 1

Resume first. A transcript for this slug (grills/<date>-<slug>.md or its -2, -3… form below, any date; newest wins, by date then suffix) with Status: in progress or abandoned after Q<n> → reopen it, never rewrite its header: pass its Q blocks to grilling as settled context (don't re-ask them), re-ask the questions it lists as unanswered, continue Q<n> and D<k>, set Status back to in progress. Otherwise write the header (today's file for this slug already exists, settled → add -2, -3… to the slug, never overwrite):

# Grill: <topic>

Date: <local date> · Subject: <doc path or one line> · Register: <decisions doc path> · Status: in progress

## Rounds

Per answer, before the next question

Transcript, per answered question. Q<n> counts across the whole session: continue the numbering across rounds, unique within the file. The heading is exactly ### Q<n> so its anchor is the lowercase #q<n> (a title in the heading would change the anchor); the title goes on the first line:

### Q<n>
- Topic: <title>
- Asked: <question + options, one line>
- Recommended: <your rec>
- Answer: <user's choice, verbatim if short; delegated → "(delegated) <rec>">
- Why: <user's reason, else the recommendation's rationale marked "(rec)">

Register, one row per settled question, continuing the D<k> numbering (an ADR folder: the next ADR number, cited ADR <n>). Source is the transcript path, relative to the register, plus the lowercase anchor (grills/2026-10-16-sync-plan.md#q3 from skilleddocs/decisions.md). Called from a wayfinder ticket, name the ticket in Source too (ticket <link>; grills/<file>#q<n>):

| D<k> | <local date> | <decision> | <why> | grills/<file>#q<n> |

New register header:

# Decisions

Append-only register of decisions from grilling sessions, consult-plan reviews, and orchestrator runs. Reverse a decision with a new row; old rows stay as written.

| # | Date | Decision | Why | Source |
|---|------|----------|-----|--------|

Every answer is a decision, delegated ones too: planners, verifiers and consult-plan read the register.

Close

Frontier empty and the user confirms → Status: settled, add ## Summary (3–8 bullets) and ## Open (or "none"). Ended early → Status: abandoned after Q<n> (n = last answered) and list asked-but-unanswered questions under ## Unanswered (a resume re-asks them). Print both paths; say they're uncommitted unless the user asks to commit, and that a tracked register's rows need their transcript committed with them or the Sources dangle.

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.