mcpField
G
grill-me-with-docs

grill-me-with-docs

github.com/rjminshaDevelopment070 installsUpdated 4d ago
Repository

About grill-me-with-docs

A skill for interviewing developers about domain terms, system boundaries, and documentation decisions. It cross-checks claims against the codebase and updates the project glossary and system map during the session.

Key features

  1. 01Ask focused questions across decision branches
  2. 02Maintain the LANGUAGE.md domain glossary
  3. 03Maintain the MAP.md system map
  4. 04Identify decisions worth documenting
  5. 05Check for existing interfaces before adding new ones

Use cases

  1. 01Clarify project terminology and record it inline
  2. 02Map component locations and system connections
  3. 03Check whether a proposed module or CLI command duplicates existing work

Install

Copy the command and run it in a terminal.

$npx skills add https://github.com/rjminsha/skills --skill grill-me-with-docs

After installing, type/grill-me-with-docsin Claude Code to confirm it loaded.

Bundled files

References and scripts installed along with SKILL.md.

DECISIONS.md
LANGUAGE-FORMAT.md
MAP-FORMAT.md
REFERENCE.md

Details

From the project README

Use the grill-me skill to conduct the interview session, focused on domain terminology, system boundaries, and documentation decisions. If grill-me is not available, conduct the interview directly using the fallback instructions in REFERENCE.md.

Used skills

SkillIf not available
grill-meInterview the human directly: ask one relentless question at a time, walk each branch of the decision tree, provide a recommended answer for each, and explore the codebase when a question can be answered by it

Domain awareness

This skill builds and maintains three living documents:

FilePurpose
LANGUAGE.mdCanonical domain terms — what things mean. Glossary only, no implementation details.
MAP.mdSparse system map — where things live and how they connect. Gives agents and developers with limited context a reference to the whole system.
docs/decisions/Hard-to-reverse, surprising, or genuine trade-off decisions.
File structure
/
├── LANGUAGE.md          ← domain glossary
├── MAP.md               ← system map
├── docs/
│   └── decisions/
│       └── 0001-*.md
└── src/

Create files lazily — only when you have something to write. See REFERENCE.md for multi-context structure.

During the session

Challenge terms against LANGUAGE.md, sharpen fuzzy language, discuss concrete scenarios, and cross-reference claims with the code. See REFERENCE.md for detailed session behaviours.

Update LANGUAGE.md inline

When a term is resolved, update LANGUAGE.md right there. Don't batch these up — capture them as they happen. Use the format in LANGUAGE-FORMAT.md.

LANGUAGE.md is a glossary and nothing else — no implementation details, no decisions.

Validate new interfaces

Before any new component, module, UI, API surface, or CLI command is created:

  1. Check MAP.md for an existing equivalent
  2. Check the codebase for an existing equivalent
  3. If one exists: "This looks like it already exists as X — is creating a new one intentional?"
  4. If confirmed, ask what it does that the existing one does not, then update MAP.md

Use the format in MAP-FORMAT.md.

Update MAP.md inline

Update MAP.md immediately when any of the following happen:

  • A new component, UI, or related system is identified or confirmed
  • A validation check reveals an entry in MAP.md needs correcting
  • End of session — review MAP.md for anything discussed but not yet captured
Offer decisions sparingly

Only offer a decision when it is hard to reverse, surprising without context, and the result of a real trade-off. See DECISIONS.md for the full criteria and format.

At the end of the session

Print the resulting file structure showing every file created or updated:

/
├── LANGUAGE.md                               ← updated
├── MAP.md                                    ← updated
├── docs/
│   └── decisions/
│       └── 0001-example-decision.md          ← created
└── src/

Comments

Markdown supported