LongTable — AI skill for Claude Code

LongTable

LongTable is a researcher-centered workspace for long scholarly projects. It helps a researcher keep research questions, construct boundaries, method choices, evidence standards, authorship judgments, and open tensions traceable across AI-assisted sessions.

LongTable is not a chatbot replacement and not a prompt collection. It is a small research workspace system that runs through a CLI, provider skills, optional MCP state and checkpoint transport, and panel collaboration records.

LongTable also exposes a narrow Codex hard-stop guard: unresolved Research Specification blockers can block session stop until the researcher decides, clears, or explicitly defers them. Product, setup, documentation, release, and hook-engineering prompts are not hard-stop blockers by default.

Codex, Claude Code, and other providers remain the execution environments. LongTable provides the durable research state, generated provider skills, Researcher Checkpoints, and command-line tools that keep important decisions from disappearing into chat history.

Core Workflow

Use one shell setup step and one provider-native research-start step.

npm install -g @longtable/cli
longtable setup --provider codex
cd ""
codex

Inside the provider, start the project with:

$longtable-start

Run pressure interviews with the LongTable grilling surface:

$longtable-interview

Required routing:

• $longtable-start is the research-start surface.
• $longtable-interview is the grilling interview surface.
• It does not require a Research Specification before asking the next pressure question.
• It frames each turn as Tension: followed by one Pressure question: and waits for a direct answer.
• If only a First Research Shape exists, LongTable continues through $longtable-start until a Research Specification is created or explicitly kept open.

`longtable start` remains a shell fallback for scripts and automation. It is not the m