Integrations

TerseForge

Use the local TerseForge CLI. Keep code quality and verification above token reduction.

Activate in the current project

1. Run terseforge --version. If unavailable, explain that the CLI must be installed; do not install remote software silently.
2. Identify the current host as codex, claude, or gemini.
3. If terseforge.config.json is absent, run terseforge init --preset <mode>, using an explicitly requested mode or safe by default.
4. If configuration exists, preserve its current mode unless the user explicitly requests another. Change an explicit mode with terseforge mode <mode>.
5. Run terseforge skill install --agent <host> --scope project so future sessions in this repository can discover the skill.
6. Run terseforge doctor. Report the mode, installed skill path, and diagnostic result briefly.

Never use init --force merely to change modes. Never overwrite an existing foreign skill or instruction file.

Work with TerseForge

• Use terseforge map and targeted terseforge context before broad file reads.
• Use terseforge exec -- <command> [args...] for noisy commands.
• Retrieve omitted content with terseforge output <run-id> whenever more evidence is needed.
• Preserve code, diffs, commands, paths, diagnostics, and security findings exactly.
• Run terseforge check before claiming completion. Required failures block verification.
• Use terseforge clean --dry-run to preview expired local state before explicit cleanup.

safe, lean, and ultra control visible context and logs, not code length or reasoning quality. Prefer safe for risky or unfamiliar work.

Inspect or stop

• For status, run terseforge doctor and terseforge skill status --agent <host> --scope project.
• To stop using TerseForge for the current task, stop invoking its commands and say that project files remain unchanged.
• Before removing persistent configuration or skill files, list the exact paths and obtain confirmation. terseforge clean removes only expired runtime state, not configuration or skill files.