grill-self

Resolves planning and design decision branches through autonomous investigation and self-questioning.

1|Updated Aug 24, 2026
One-click install
npx skills add https://github.com/mjun0812/skills --skill grill-self-mjun0812
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: grill-self
Source: https://github.com/mjun0812/skills/tree/main/skills/planning/grill-self
Command: npx skills add https://github.com/mjun0812/skills --skill grill-self-mjun0812

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Planning and design work often stalls on unresolved decision branches, and asking the user for every judgment call breaks flow. This Skill drives the agent to investigate the codebase, conventions, and docs itself, resolve each decision point one by one, and deliver a complete decision log without asking the user any questions. ## Core Features & Use Cases - Autonomous Branch Resolution: Walks the design tree depth-first, resolving one decision at a time through investigation, proposal, devil's-advocate rebuttal, and adoption. - Evidence-Based Decisions: Requires examining the codebase, existing conventions, and documentation before each decision instead of guessing. - Decision Log Output: Produces a final table of every decision with its rationale and confidence level (high/medium/low), flagging low-confidence items for review. - Use Case: When drafting an implementation plan for a new feature, ask the agent to self-grill the design so every open question—naming, data model, error handling, compatibility—is decided with documented reasoning before coding starts. ## Quick Start Use the grill-self skill to work through all open design decisions for this feature plan without asking me any questions, then show me the decision log.

Frequently Asked Questions about grill-self

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I get an AI agent to make design decisions without asking me questions?

Use a self-questioning planning workflow where the agent investigates the codebase and docs, then resolves each decision branch itself. This Skill walks the design tree depth-first and records every decision with rationale and confidence instead of blocking on user input.

How to resolve open questions in a software design plan automatically?

The agent enumerates each decision point, gathers evidence from code and conventions, proposes a recommendation, argues against it, then adopts it with a confidence level. Value judgments that only the user can make are provisionally adopted at low confidence rather than left open.

What output does an autonomous planning review produce?

It produces a decision log table listing each issue, the decision made, its rationale, and a confidence rating of high, medium, or low. Low-confidence items are relisted as needing confirmation, with a note on the impact if the decision is reversed.

When should I not use autonomous self-questioning for planning?

Avoid it when decisions depend on business priorities or user preferences that cannot be inferred from the codebase, since those become low-confidence provisional choices. It also prioritizes depth over breadth, so it is less suited for quick high-level surveys.

Does the agent ever ask the user questions during this process?

No. The workflow explicitly forbids asking the user anything. Undecidable value judgments are provisionally adopted with low confidence and flagged for later review in the final decision log.