Agent Skills

Jumpspace is most useful when agents know to use it first, then fall back to code reading when needed.

Install Skills

npx jumpspace add-skill --codex
npx jumpspace add-skill --claude
npx jumpspace add-skill --all

The installer is additive and non-destructive. It creates missing files or updates clearly marked Jumpspace-managed blocks. It should not overwrite user-authored guidance outside those markers.

What The Skill Tells Agents

The generated guidance tells agents to:

• run Jumpspace discovery before broad repository reads
• treat ask as evidence, not authority
• use JSON schemas before scripting against command output
• use work for approved implementation
• use next for plan steps
• record step evidence
• use verify to earn verified state
• run scan, audit, and doctor after metadata mutations
• fall back to rg and direct code reads when Jumpspace evidence is missing

Useful Agent Commands

npx jumpspace find "approval flow" --json --compact
npx jumpspace context DOC-EXAMPLE-001 --json
npx jumpspace work DOC-EXAMPLE-001 --json
npx jumpspace next DOC-EXAMPLE-001 --json
npx jumpspace handoff --task DOC-EXAMPLE-001 --json
npx jumpspace last --json
npx jumpspace history --json

Check The Active Install

When an agent seems to be running an old CLI, run:

npx jumpspace release install-doctor --json

This reports the invoked binary, the first jumpspace found on PATH, package roots, versions, schema counts, workspace comparisons, stale-install warnings, and repair commands.

Human Pattern

For a small team, the ideal pattern is:

1. The human writes or approves the task block.
2. The agent uses Jumpspace to gather context.
3. The human approves the plan.
4. The agent implements the next step.
5. Jumpspace records evidence and verification.
6. CI checks drift before merge.