Skills

A skill is a markdown file that gives Claude a defined mission. Skills live in your skills folder and appear in the / menu — and optionally in the Obsidian command palette — ready to run on demand.

Skills range from simple reusable prompts to fully parameterized agentic workflows. Because Claude runs with full tool access, a skill can read and write vault notes, run shell commands, call APIs, and more. If you can describe it in a prompt, you can make it a skill.

File format

A skill file is a standard markdown file with optional YAML frontmatter:

---
category: Reviews
description: Structured weekly review
autorun: true
params:
  - id: focus
    type: input
    label: Any specific focus area?
    placeholder: Leave blank for a full review
---
Review my notes from this week. Give me:
- What I worked on
- Open threads or unresolved questions
- Suggested priorities for next week

{{focus}}

The frontmatter controls how the skill appears in the menu and how it behaves. The body is the prompt Claude receives.

Frontmatter reference

Field	Type	Default	Description
category	string	Prompts	Groups the skill under a named heading in the / menu
description	string	(none)	Short subtitle shown below the skill name
autorun	boolean	false	If true, executes immediately. If false, inserts into the chat input for review
params	array	(none)	Form fields presented to the user before the skill runs

params fields

When params is defined, BojuBot shows a modal form before running the skill. Each entry is an object with the following properties:

Property	Required	Description
id	Yes	Variable name used in in the prompt body
type	Yes	Field type — see below
label	Yes	Human-readable label shown in the form
description	No	Helper text shown below the label
placeholder	No	Placeholder text for input and textarea fields
options	Yes (dropdown, checkboxes)	Array of string values
default	No	Pre-filled value
validations.required	No	If true, blocks submission when the field is empty

Field types

Type	Description
input	Single-line text field
textarea	Multi-line text field
dropdown	Select from a fixed list of options
checkboxes	One or more boolean toggles — result is a comma-separated string
obsidianmd_note	Vault note picker — fuzzy search over all vault notes; injects the full note content as an attachment (same as @-mention). Alias: note (legacy, still supported)

Variable interpolation

Use {{id}} in the prompt body to reference field values. After submission:

• Non-note fields are interpolated inline.
• Note fields are added as context attachments (shown as badges above the input, just like @-mention). The {{id}} token is stripped from the body — the note content arrives via the attachment, not inline text.
• Unresolved tokens (optional fields left blank) are stripped cleanly. No visible {{placeholder}} tokens appear in the output.

Handling optional parameters

Interpolation only strips the {{id}} token itself — any static label text you wrote around it stays. If an optional textarea param is left blank, a body like:

**Additional instructions:** {{extra_notes}}

leaves a dangling **Additional instructions:** with nothing after it. It's harmless, but noisy, and can read like a cut-off instruction to Claude.

Wrap optional parameters in XML-style tags instead:

<extra_instructions>
{{extra_notes}}
</extra_instructions>

Claude treats XML-like tags as semantic containers, so this pattern handles both cases cleanly:

• Empty — Claude sees a clearly-bounded empty block and ignores it.
• Populated — the user's content is unambiguously scoped, so it can't blend into the skill's own surrounding instructions.

This matters most for free-text fields (input, textarea) where the user might paste in dense or instruction-like text. Use it for any optional parameter, not just the ones that happen to be empty most of the time.

Keep internal protocol details out of skill bodies. Skills intended for general use should describe actions in plain language ("open the file in Obsidian") and let Claude's system context handle the mechanics — don't embed raw UI Bridge trigger syntax directly in a skill prompt body.

Execution modes

autorun	params	Behaviour
false	none	Inserts prompt body into chat input for review
true	none	Fires immediately; shows Running: <name> in chat
false	defined	Shows form → inserts filled prompt into chat input
true	defined	Shows form → fires immediately on submit

Skills folder

Default: _BojuBot Skills/ — at your vault root. This keeps skills in your vault, making them easy to edit and sync with Obsidian Git.

Custom: set Settings → BojuBot → Skills folder to:

• A vault-relative path (e.g. _skills) — keeps skills in your vault, synced with Obsidian Git
• An absolute path — for a shared folder outside the vault

Skills reload each time you open the / menu, so changes take effect immediately with no restart.

Examples

Simple autorun

---
category: Writing
description: Summarize the active note
autorun: true
---
Summarize the currently active note concisely, preserving key decisions and open questions.

Weekly review with optional focus

---
category: Reviews
description: Structured weekly review
autorun: true
params:
  - id: focus
    type: input
    label: Focus area (optional)
    placeholder: Leave blank for a full review
---
Review my notes from this week and give me:
- What I worked on
- Open threads or unresolved questions
- Suggested priorities for next week

{{focus}}

Bug report with form

---
category: GitHub
description: File a bug report on any repo
autorun: true
params:
  - id: repo
    type: input
    label: Repository
    placeholder: owner/repo
    validations:
      required: true
  - id: title
    type: input
    label: Bug title
    validations:
      required: true
  - id: severity
    type: dropdown
    label: Severity
    options: [low, medium, high, critical]
    default: medium
  - id: body
    type: textarea
    label: Description
    placeholder: Steps to reproduce, expected vs actual behavior…
---
File a GitHub issue in {{repo}}.

Title: {{title}}
Severity: {{severity}}

{{body}}

Note summarizer using vault picker

---
category: Writing
description: Summarize any vault note
autorun: true
params:
  - id: target_note
    type: obsidianmd_note
    label: Note to summarize
    validations:
      required: true
---
Summarize the attached note concisely, preserving key decisions and open questions.

The picked note is injected as a context attachment — Claude reads it exactly as it would read an @-mentioned note.

Programming agent

---
category: Code
description: Implement a feature in this repo
autorun: true
params:
  - id: feature
    type: textarea
    label: What should I build?
    placeholder: Describe the feature in plain language…
    validations:
      required: true
  - id: context_note
    type: obsidianmd_note
    label: Relevant design note (optional)
---
You are acting as a senior software engineer working in this vault's codebase.

Task: {{feature}}

Read the relevant files, understand the existing architecture, implement the feature, and write a brief summary of what you changed and why.

Ctrl+P integration

Enable Settings → Register skills as Ctrl+P commands to expose every skill as an Obsidian command palette entry, prefixed Skill: ….

After adding or removing skill files, run BojuBot: Reload skills from the palette to sync. You can also assign hotkeys to individual skills via Settings → Hotkeys — search for Skill: to find them all.

See the Skills API section of the Commands reference for details.