Skip to content
Liner

Liner Docs

Liner is a local CLI and terminal UI for turning focused research sources into reusable AI context.

You give it URLs, transcripts, PDFs, local files, or pasted text. It saves those materials, compiles the useful research into Markdown, and writes instruction files future AI sessions can read.

A Liner Project is the durable local folder that stores that work. It contains a curated Mixtape plus LINER.md, a root SKILL.md, and liner.yaml.

What Liner Is For

Section titled “What Liner Is For”

Use it when source choice changes the quality of the AI’s work. It fits focused jobs like reviewing a native iOS flow, comparing API migration patterns, preparing a hiring rubric, or grounding a strategy review in a specific body of evidence.

The workflow helps you:

  • frame the task and desired output
  • gather and evaluate source material
  • write a summary the AI should inherit
  • fetch source content into local Markdown files
  • create instruction files for future AI sessions
  • safely maintain Project and Source state over time
  • package the folder for reuse or sharing

The Core Workflow

Section titled “The Core Workflow”
  1. Create a Liner Project and define its Job to Be Done in Job Story form.
  2. Add any material you already have through the Source Inbox, review the active Sources, and save that selection.
  3. Clarify Job so the desired outcome, situation, and boundaries are explicit.
  4. Build the corpus with your configured local AI runner.
  5. Review the draft Source set and accept the Sources you want to keep.
  6. Review the Synthesis separately. Record approval when it accurately represents the accepted Sources, or edit it before approval.
  7. Liner starts Compile after Synthesis approval and writes the Mixtape.
  8. Create the Operating Layer files in the terminal UI.
  9. Give the Liner Project to your AI tool.
  10. Optionally install a Maintenance Adapter so Codex or Claude can maintain it through Liner’s reviewed CLI workflow.
$npx linersh

Key Ideas

Section titled “Key Ideas”

Tape: the small YAML recipe. It lists metadata and source entries.

Mixtape: the research packet inside the project. In v2 projects it lives under mixtape/ and contains the tape, synthesis, extracted sources, working notes, and compiled output.

Liner Project: the full working folder. It contains the Mixtape plus AI instructions, a root skill file, status, and maintenance metadata.

Synthesis: the curator’s distilled framing. This is copied into the top of MIXTAPE.md.

Sources: web pages, YouTube transcripts, or local files that Liner fetches and stores as Markdown.

Project and Source IDs: immutable identifiers assigned by Liner. Renaming or moving a Project keeps its Project ID. Editing Source metadata or its locator keeps its Source ID; replacing the meaning of a Source creates a successor ID and records lineage.

Instruction files: LINER.md, root SKILL.md, and liner.yaml. These files tell future AI sessions how to use the saved research and what state to trust. In the TUI, the action that creates them is called Create Operating Layer.

Optional Maintenance Adapter: a thin Codex or Claude skill that routes Project and Source changes through the installed CLI. It previews a Project Change Set, waits for Curator approval when required, and leaves a Change Receipt after apply. It does not replace the Project Skill.

Project Snapshot, Change Set, and Change Receipt: the safe maintenance sequence. A Snapshot is read-only current state, a Change Set is an exact write-free plan for one Project revision, and a Receipt is the durable redacted record of the applied plan.

Next

Section titled “Next”

Start with Install, walk through Build a Project, then see Maintain a Project with an Agent when you want Codex or Claude to update it safely.