deep-research-interview

deep-research-interview is a command for coding agents from axiomantic/spellbook. It costs 36 tokens per session (2,304 once invoked), scanned A, original, MIT.

An interview command that turns a broad research request into a structured Research Brief. The brief defines the subject, assumptions, questions, and boundaries for later research.

In plain words
What is it for?
Use it to clarify complex research requests before gathering sources or producing findings.
Why use it?
It exposes ambiguity and hidden assumptions early, reducing the risk of researching the wrong person, organization, date, or question.

Command

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add commands/axiomantic/spellbook/deep-research-interview
Clone the repo
git clone --depth 1 https://github.com/axiomantic/spellbook

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

agentmods badge for deep-research-interview

README.md
[![agentmods](https://agentmods.dev/badge/commands/axiomantic/spellbook/deep-research-interview.svg)](https://agentmods.dev/commands/axiomantic/spellbook/deep-research-interview)
Your own site
<a href="https://agentmods.dev/commands/axiomantic/spellbook/deep-research-interview"><img src="https://agentmods.dev/badge/commands/axiomantic/spellbook/deep-research-interview.svg" alt="Measured on agentmods" height="20"></a>
Per session 36 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 2,304 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5 $0.00036 $0.02304
Opus 5 $0.00018 $0.01152
Sonnet 5 $0.00007 $0.00461
Haiku 4.5 $0.00004 $0.00230

Measured 4d ago against content hash 8537e8444ffb, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

deep-research-interview scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 4d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

commands/deep-research-interview.md · 272 lines

How it starts

The opening of the file, as written. The whole thing — 272 lines — stays where its author put it; the contents beside it link to each section on GitHub.

MISSION

Transform a raw research request into a Research Brief by surfacing implicit assumptions, conducting a structured interview across 5 categories, and producing a brief that serves as the contract for all subsequent research phases.

Invariant Principles

  1. Brief is the contract - Every item in the Research Brief drives all subsequent phases; nothing outside it gets researched, nothing inside it gets skipped
  2. Assumptions are liabilities - Every unstated assumption is a wrong-target risk; surface and verify each one before locking scope
  3. Disambiguation before depth - A research effort aimed at the wrong entity is worse than no research; resolve identity first
  4. Interview is adaptive - Stop when criteria are met, not when questions run out; never ask what you already know

Step 1: Prompt Improvement

Before interviewing, analyze the raw request for implicit assumptions and disambiguation needs.

1.1 Assumption Extraction

For each factual claim in the user's request, classify and surface:

Claim Type Questions to Surface
Date/Time What is the source? How precise? Could it be approximate?
Name/Entity Known variants? Is this the common name in context?
Location/Scope Has jurisdiction or boundary changed over time?
Relationship What evidence supports this link?
Institution Does it still exist? Have records been transferred?
Record/Artifact Type Does this exist for this period or context?

1.2 Disambiguation Need Identification

Run these 5 checks against the request:

  1. Name Frequency - Is this a common name in context? Flag for disambiguation if yes.
  2. Generational Check - Same-named relatives, versions, or editions? Require temporal anchoring.
  3. Spelling/Naming Stability - Inconsistent conventions across sources? Generate search variants.
  4. Jurisdictional/Scope Stability - Boundaries changed over time? Identify all relevant scopes.
  5. Record Type Existence - Does the requested artifact actually exist for this period/context? Identify alternatives if not.

Read the full file on GitHub · 272 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 4d ago First seen · 272 lines · 36 tokens per session scan A 8537e8444ffb

Subscribe to this mod's changes

deep-research-interview is a command published in the GitHub repository axiomantic/spellbook (10 stars, last pushed yesterday), licensed MIT. It adds 36 tokens to every session and 2,304 once invoked, about $0.0002 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.