Documentation
¶
Overview ¶
Package guides ships brainjar-authored markdown content as Claude Code SKILL.md files. Two transports serve the same registry:
- MCP — `guide_list` / `guide_show` so an agent driving brainjar can pull a guide on demand.
- Disk emit — `brainjar sync` writes every guide to the active platform's skills directory so Claude Code's native loader picks them up via description-match.
Guides are server-shipped, embedded in the binary. Updates ride the binary: `brainjar upgrade` swaps the guides the same instant it swaps the tools they document. There is no SQLite seed and no user authoring path; user content uses souls/personas/rules/procedures/ skills.
Index ¶
Constants ¶
const MarkerFile = skillemit.MarkerFile
MarkerFile sits inside every brainjar-emitted skill directory so the prune step knows which dirs we own. User-authored skills share the same parent directory but lack this file, so they survive every sync.
Aliased from internal/platform/skillemit so callers in this package keep their existing import surface.
Variables ¶
This section is empty.
Functions ¶
func AsSkillEmits ¶
AsSkillEmits returns every guide as a platform.SkillEmit suitable for passing to Platform.EmitSkills. Raw is preserved so the on-disk SKILL.md bytes match the embedded source.
Types ¶
type Guide ¶
type Guide struct {
Slug string `json:"slug"`
Description string `json:"description"`
Version int `json:"version"`
Checksum string `json:"checksum"`
Scope Scope `json:"scope"`
Body string `json:"body"`
// Raw is excluded from JSON — it's only used by the emit step.
Raw []byte `json:"-"`
}
Guide is a single embedded guide. Each guide is shipped on disk as a Claude Code SKILL.md file at `content/<slug>/SKILL.md`. The slug is the directory name, validated against the `name:` field in the SKILL.md's YAML front-matter so the on-disk name and the in-binary identity stay in sync.
Raw is the byte-for-byte file content (frontmatter + body) used by the disk-emit step; Body is the post-frontmatter markdown returned by `guide_show` and `brainjar guide show`.
type Registry ¶
type Registry struct {
// contains filtered or unexported fields
}
Registry holds the parsed embedded guides keyed by slug.
func Default ¶
func Default() *Registry
Default returns the singleton registry over the embedded guides. Parse failures are fatal — the embed ships inside the binary, so a malformed file is a build bug, not a runtime one.
func Load ¶
Load walks dir within fsys for `<slug>/SKILL.md` files and parses each into the registry.
func (*Registry) All ¶
All returns every guide in slug order, including bodies and raw bytes. Used by the emit step.
func (*Registry) AllAtScope ¶
AllAtScope returns every guide whose Scope == s, in slug order. Used by the emit step to split per-scope target dirs.
func (*Registry) Get ¶
Get returns the guide with the given slug. Unknown slug returns a CodeNotFound app error so callers (MCP tools, CLI) surface the right envelope.
func (*Registry) IsReservedSlug ¶
IsReservedSlug reports whether slug names a built-in guide. Skill save paths reject reserved slugs so a workspace skill can't silently overwrite a shipped guide's SKILL.md on next sync.
type Scope ¶
type Scope string
Scope selects where a guide's SKILL.md is emitted on `brainjar sync`. ScopeProject lives under `<projectRoot>/.claude/skills/` — only Claude Code sessions in that project see the skill. ScopeUser lives under `~/.claude/skills/` — every Claude Code session sees it.
type Summary ¶
type Summary struct {
Slug string `json:"slug"`
Description string `json:"description"`
Version int `json:"version"`
Checksum string `json:"checksum"`
Scope Scope `json:"scope"`
}
Summary is the list view returned by guide_list and brainjar guide list. No body, no raw — just the lookup metadata.
Source Files