Add /grill-me planning interview skill

.claude/skills/grill-me/README.md

@@ -0,0 +1,52 @@
+# Grill Me (`/grill-me`)
+
+> **A relentless interview about your plan or design — one question at a time, walking down the decision tree.**
+
+`/grill-me` is the simplest skill in the kit. It does one thing: it interviews you about whatever plan, design, or proposal you're working on, asking one focused question at a time, providing its recommended answer for each, and continuing until every branch of the decision tree has been resolved.
+
+## What it does
+
+When you invoke `/grill-me [optional context]`, Claude takes the position of a thoughtful collaborator who refuses to let you skip past unresolved decisions. It walks down the design tree, surfacing dependencies between choices and forcing each one to be made explicitly before moving on.
+
+For each question:
+- The question is asked one at a time (never a list — each answer changes what comes next)
+- Claude provides its **recommended answer** with reasoning, so you can agree, disagree, or redirect
+- If the question can be answered by reading the codebase, Claude reads the codebase instead of asking you
+- The conversation continues until shared understanding is reached
+
+## When to use it
+
+- **Before implementing** something nontrivial — when there are too many open design questions to start writing code yet
+- **When a plan feels vague** — invoke `/grill-me` to surface the decisions you've been glossing over
+- **When you want a stress test** — to check whether your design holds up under interrogation
+- **For audits** — pointing it at a directory, a skill, a paper, or any artifact and asking for a structured review
+
+## Usage
+
+```
+/grill-me                              # generic — Claude asks what to grill you about
+/grill-me [the plan / design / topic]  # interrogation begins immediately
+/grill-me audit my custom skills before sharing with colleagues
+/grill-me whether this identification strategy holds up
+```
+
+The argument is free-form — describe what you want grilled, and Claude will treat it as the subject of the interview.
+
+## Philosophy
+
+The skill exists because most plans fail at points the author already noticed but didn't want to think about. A relentless interviewer who walks the decision tree and refuses to skip ahead surfaces those points before they become bugs, lost work, or wasted code.
+
+The "one question at a time" rule is structural, not stylistic. A list of questions invites batch answers; batch answers invite hand-waving. A single question, with a single recommended answer, demands a single explicit decision.
+
+## Files in this skill
+
+- [`SKILL.md`](SKILL.md) — the skill definition. Very short. The skill's behavior is the protocol, not a long instruction set.
+
+## Related skills
+
+- **`/blindspot`** — also an interrogation tool, but pointed at empirical *output* (figures, tables) rather than plans. `/grill-me` walks the design tree before work starts; `/blindspot` walks the perception tree after results exist.
+- **`/referee2`** — a structured implementation audit. `/grill-me` is open-ended; `/referee2` runs a fixed five-audit checklist.
+
+## Acknowledgments
+
+The `SKILL.md` is taken verbatim from [Matt Pocock's grill-me skill](https://github.com/mattpocock/skills/tree/main/skills/productivity/grill-me) — full credit for the skill's behavior and design goes to him. The only contribution here is this README, which adds a human-readable explanation of what the skill does, when to use it, and how it relates to the other skills in this collection.

.claude/skills/grill-me/SKILL.md

@@ -0,0 +1,10 @@
+---
+name: grill-me
+description: Interview the user relentlessly about a plan or design until reaching shared understanding, resolving each branch of the decision tree. Use when user wants to stress-test a plan, get grilled on their design, or mentions "grill me".
+---
+
+Interview me relentlessly about every aspect of this plan until we reach a shared understanding. Walk down each branch of the design tree, resolving dependencies between decisions one-by-one. For each question, provide your recommended answer.
+
+Ask the questions one at a time.
+
+If a question can be answered by exploring the codebase, explore the codebase instead.

skills/README.md

@@ -21,6 +21,7 @@ This directory contains documentation, methodology, and example output for the s
 | [**TikZ Audit**](tikz/) | `/tikz` | **A repair tool, not a safety net.** Finds and fixes residual visual collisions in TikZ figures using measurement, not intuition — six-pass protocol covering Bézier curve depths, edge-label gap calculations, boundary clearances, and cross-slide consistency. Catches what `pdflatex` misses. But it cannot reliably fix diagrams that were never built with measurement in mind. The upstream defense is `/beautiful_deck` Step 4.4, which writes safe TikZ from the start. `/tikz` is the downstream check. [See documentation →](tikz/) |
 | [**Split-PDF**](split-pdf/) | `/split-pdf` | Downloads and deep-reads academic PDFs without crashing the session. Uses the PDF in place (no centralized `articles/` folder), splits into 4-page chunks in a `_build/` directory, reads in batches of ~12 pages, writes structured notes, and saves a persistent `_text.md` extraction so future invocations skip re-reading. When called by another skill, reads inside a subagent to prevent context bloat. [See full walkthrough →](split-pdf/) |
 | [**Bibcheck**](bibcheck/) | `/bibcheck` | Many-agent bibliography audit. Spawns one narrow-focus agent per citation (or one specialist per field) to verify each `.bib` entry against canonical sources — DOI, journal landing page, author working paper. Catches the silent errors a single-agent audit misses as attention decays across long lists: mixed-up entries (title of paper A with authors of paper B), wrong years, journal misattributions. Per-field mode launches each specialist as an isolated `claude -p` subprocess so they cannot peek at each other's conclusions. Outputs a `bibcheck_report.md` and a drop-in `corrected.bib`. [See documentation →](bibcheck/) |
+| [**Grill Me**](grill-me/) | `/grill-me` | A one-question-at-a-time planning interview. Stress-tests plans and designs before implementation, gives a recommended answer for each decision point, and keeps going until the decision tree is resolved. [See documentation →](grill-me/) |
 | [**New Project**](newproject/) | `/newproject` | Scaffolds a new research project with standard directory structure, CLAUDE.md template, and documented README. [See documentation →](newproject/) |
 | [**New Book**](newbook/) | `/newbook` | Scaffolds a book-shaped project: `memoir`-based LaTeX skeleton, Palatino body, Gov 2001 palette, voiced-sidebar `\voice{<name>}{<text>}` callouts, one chapter per file, bibliography stub, `CLAUDE.md` with voice and lineage rules, `% SUBSTACK MAP:` placeholders in each chapter, and a chapter-per-file structure that converts cleanly to HTML later. Parallel to `/newproject`. This is the skill that produced *AI Agents and the Research Worker*. [See documentation →](newbook/) |
 

skills/grill-me/README.md

@@ -0,0 +1,35 @@
+# Grill Me (`/grill-me`)
+
+> **A relentless interview about your plan or design — one question at a time, walking down the decision tree.**
+
+`/grill-me` interviews you about whatever plan, design, or proposal you're working on, asking one focused question at a time, providing its recommended answer for each, and continuing until every branch of the decision tree has been resolved.
+
+## What it does
+
+When you invoke `/grill-me [optional context]`, Claude takes the position of a thoughtful collaborator who refuses to let you skip past unresolved decisions. It walks down the design tree, surfacing dependencies between choices and forcing each one to be made explicitly before moving on.
+
+For each question:
+- The question is asked one at a time
+- Claude provides its recommended answer with reasoning
+- If the question can be answered by reading the codebase, Claude reads the codebase instead of asking you
+- The conversation continues until shared understanding is reached
+
+## When to use it
+
+- Before implementing something nontrivial
+- When a plan feels vague
+- When you want a stress test
+- For open-ended audits of a directory, skill, paper, or artifact
+
+## Usage
+
+```
+/grill-me
+/grill-me [the plan / design / topic]
+/grill-me audit my custom skills before sharing with colleagues
+/grill-me whether this identification strategy holds up
+```
+
+## Acknowledgments
+
+The `SKILL.md` is taken verbatim from [Matt Pocock's grill-me skill](https://github.com/mattpocock/skills/tree/main/skills/productivity/grill-me). Full credit for the skill's behavior and design goes to him. The contribution here is packaging and documentation for this skills collection.