Shared guides

Guides about the catalogue itself — installing it, upgrading it, seeing what each agent tool supports, authoring new skills — rather than any one pack. Pack-specific guides live under each pack’s own home; these are the cross-cutting ones.

How-to

• Install agentbundle from a clone — the fastest route to the CLI.
• Install a user-scope pack into Codex — and what travels with you across projects.
• Install a user-scope pack into Kiro — the Kiro-specific layout.
• Preview an install or upgrade with --dry-run — see every file that would change before it changes.
• Upgrade an installed pack — pull new pack content without clobbering your edits.
• Author a skill — write a new skill for any pack, to the catalogue’s standards.
• Author a browser-automation skill — persistent-profile auth, bearer token interception, probe files, and ui-patterns.md maintenance for skills that drive a browser.
• Choose a tracker integration — pick the right brief-intake skill for your tracker (GitHub, Linear, Jira, Jira Align, or none).
• Start, remember, inspect, or refresh repository work — use the shared content-based front door and keep local and remote writes explicit.
• Create an external catalogue — scaffold a catalogue outside this repository, validate it, and publish it via CI.

Reference

• agentbundle reference — install the CLI, install a pack, configure the default adapter, source-resolution chain, environment variables.
• Catalogue CI contract — provider-neutral contract for validation, packaging, publication ordering, exit codes, and responsibility boundaries in a catalogue CI pipeline.
• Adapter support matrix — which primitives each agent tool receives, and where it degrades.
• Tracker vocabulary — how brief and spec levels map across GitHub, Linear, Jira, and Jira Align; skill routing table.
• Work-intake routing and lifecycle — exact cross-profile routes, authority modes, dispatch rules, refresh states, and migration boundary.
• Output rendering directives — the canonical directive catalog: which shape to declare in ## Output rendering (Table, Status list, Severity list, Diagram, etc.) and when to omit.
• Skill UX patterns — craft-depth companion to output-rendering: column alignment rules, truncation limits, persistent command bar, delete-gate box, card format for review flows.
• Skill script conventions — CLI flag conventions, usage docblocks, shortcut IDs, shared-libs pattern, idempotent setup, pack-config API.

Explanation

• The three loops as a system — how discovery, build, and release compose into a complete operating model, and why each loop is a peer supervisor.
• Install routes — the four ways pack content reaches a repo, and where each lands.
• The pack catalogue — what a pack is, how the catalogue is composed, and how you build your own.
• The file-safety contract — why your edits are never silently overwritten.
• Shaping a new engagement — how a product vision, a product strategy, and an architecture concept co-shape each other at the start of a new engagement.
• How work records divide responsibility — why source evidence, canonical artifacts, workspace state, processors, and the migration ledger have separate jobs.

---

Each quadrant subdirectory also carries a short writing-rules README (tutorials · how-to · reference · explanation) — what belongs in that kind of guide, for anyone authoring one.