# Gantry Mode

> Configure Gantry guidance globally or for an active task. Use when the engineer invokes /gantry-mode, asks to change guidance mid-session, or wants guided, collaborative, or concise Gantry interactions.

- Skill: `karthikcsq/gantry-mode` (Agent Skill, multi-file: 4 files)
- Install (CLI): `npx skillmds@latest add karthikcsq/gantry-mode`
- Raw SKILL.md: https://api.skillmd.com/api/skills/karthikcsq/gantry-mode/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Productivity
- Author: karthikcsq (https://skillmd.com/u/karthikcsq)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/karthikcsq/gantry-mode

---


# gantry-mode

Set the default interaction style used when Gantry creates a new task, or change an existing task immediately. This changes explanation depth and pseudocode chunking only; it never weakens Gantry's approval gate or design rigor.

Store configuration at `~/.gantry/config.json` (`$HOME/.gantry/config.json`, using the platform's real user-home directory). Use the bundled script for every read and write:

```text
node <absolute-path-to-this-skill>/scripts/gantry-mode.mjs [guided|collaborative|concise|reset] [slug] [--root <project-root>]
```

## Arguments

Accept a mode and an optional Gantry slug:

- no argument — report the current default and list the three valid levels.
- `guided`, `collaborative`, or `concise` — set the user-level default for future tasks.
- `<mode> <slug>` — update `.gantry/<slug>.diff.md` and apply the level to that task immediately without changing the default. Applying a level to an active task means re-rendering the doc's unresolved givens, forks, and open items in the new style inside `.gantry/<slug>.md`; it does **not** mean re-explaining the existing design in chat. Keep the chat reply to the script's confirmation plus a one-line pointer to what changed in the editor.
- `reset` — remove the global override and return the default to `collaborative`.
- `reset <slug>` — set that task to the current global default.

Treat values case-insensitively and normalize them to lowercase. For any other value, print the valid choices and stop without changing the file.

Run the script with no argument to report status. For a task-specific change, pass the repository root with `--root` and use the filename stem from `.gantry/<slug>.md`. Relay the script's output without embellishment. If it reports malformed configuration or sidecar frontmatter, stop; never repair or overwrite it automatically.

The script preserves unrelated global configuration keys and sidecar frontmatter.

## Output

Keep responses short. Every script call — status, a default change, a task change, or a reset — lists the three available modes so the engineer always sees the choices. Relay that verbatim; do not restate or re-describe the modes yourself.

Status call:

```text
Gantry guidance: collaborative

guided — more teaching and smaller steps
collaborative — balanced pairing (default)
concise — terse, fluency-assuming context
```

A write additionally reports the exact config or sidecar path and appends the same `Available modes:` block:

```text
Gantry guidance set to guided for future tasks.
Config: /home/you/.gantry/config.json

Available modes:
guided — more teaching and smaller steps
collaborative — balanced pairing (default)
concise — terse, fluency-assuming context
```

