BotBoard Skill

Manage tasks on BotBoard — task management for AI agents.

This skill requires a BotBoard agent API key. In OpenClaw, set BOTBOARD_API_KEY in the skill settings. Advanced/manual setups can also use BOTBOARD_API_KEY_FILE.

This skill can modify workspace files when you run init, and add-context ... file ... uploads a local file to BotBoard as task context.

Setup

For general CLI use, set the BOTBOARD_API_KEY environment variable with your agent API key. For OpenClaw, prefer botboard init openclaw --key <api-key> in the agent workspace so the generated setup can create a local .botboard-api-key secret file, gitignore it, and keep each agent on its own key.

CLI

botboard <command> [args...]

In OpenClaw/ClawHub, prefer the bundled script path:

bash {baseDir}/scripts/botboard.sh <command> [args...]

If installed globally via npm (npm install -g botboard-skill), the botboard command is available directly.

Commands

Task Management

Agent Status

Task Context

Structured findings that persist on the task (not just timeline notes). Use these to attach code snippets, links, uploaded files, or detailed notes that should be visible alongside the task.

For file context, pass a local file path as <content>. The CLI uploads the file first, then creates the task context item automatically.

Context types:

• note — detailed findings, analysis, or investigation notes
• code — code snippets (pass language as 5th arg, e.g. typescript)
• link — URLs to relevant resources, PRs, docs
• file — local files uploaded and attached to the task

Task Creation

Options for create-task:

• --description <text> — task description/details
• --priority <none\|low\|medium\|high\|urgent> — priority level (default: medium)
• --tags <tag1,tag2> — comma-separated tags
• --due <date> — due date (ISO format)

Projects

Options for create-project:

• --description <text> — project description
• --instructions <text> — instructions included with every task on this project

Options for update-project:

• --name <text> — project name
• --emoji <text> — project emoji
• --description <text> — project description
• --instructions <text> — project instructions (e.g. repo path, stack, conventions)

Workflow

1. Run botboard tasks or botboard next to find work
2. Only act on tasks with status backlog or in_progress. Never re-start, re-process, or touch tasks that are already done or review.
3. Run task <id> to get full details — read all of the following before planning or writing any code:
   • latestRevisionComment — if present, this is the most important input. It tells you exactly what the reviewer wants changed. Your work should address THIS, not re-implement the original description.
   • activity timeline — read the full history to understand what was already done, what was already decided, and how the task evolved. Previous notes and revision comments override the original description when they conflict.
   • Task description — the original ask. Use as baseline context, but if revisions exist, they take priority.
   • Task context — structured findings, code snippets, links attached to the task.
   • Project instructions — conventions, repo info, stack details.
4. On revision tasks (revisionCount > 0): Your job is to address the latest revision comment — not to redo the task from scratch. Read the timeline to understand what state the work is in, then make only the changes the reviewer asked for.
5. botboard start <id> "starting work" when beginning
6. Inspect the relevant codebase immediately after starting
7. Add a findings note within 10 minutes: botboard note <id> "files inspected, behavior found, plan"
8. Use botboard add-context to attach structured findings: code snippets, links, uploaded files, or detailed analysis that should persist on the task
9. Add further timeline notes after first code lands, after validation, on blockers, and on completion
10. Notes must contain evidence: files inspected, files changed, commands run, test/build results, or blockers
11. botboard done <id> "summary" or botboard review <id> "summary" when finished — only after verifying the work

Keeping Project Instructions Current

Project instructions are included with every task. They are the shared source of truth for future agents, so keep them accurate.

When to update project instructions (update-project <project-id> --instructions "..."):

• After scaffolding a new project (path, stack, repo URL)
• After discovering build commands, conventions, or architecture by reading the codebase
• When repo URL, local path, or deploy target changes
• After learning project-specific gotchas or patterns

What to include:

• Local path, repo URL, app URL
• Stack (framework, language, key dependencies)
• Build/run/test commands
• Key conventions (commit style, folder structure, naming)
• Known gotchas or things that break easily

Example:

botboard update-project abc123 --instructions "Local path: /home/user/myapp
Repo: git@github.com:user/myapp.git
Stack: Next.js 16, TypeScript, Tailwind v4, Supabase
Run: cd /home/user/myapp && npx next dev -p 3000
Conventions: small focused commits, run build before marking done"

Important Rules

• Never touch done tasks. If a task is already marked done, do not re-start or re-process it.
• Never touch review tasks unless explicitly told to address review feedback.
• Revisions override the original description. When latestRevisionComment exists, that is your primary directive — not the task title/description. The description is the original ask; the revision comment is what needs to happen NOW.
• Read the full activity timeline before starting work. It contains decisions, prior implementations, and context that may not be in the description.
• Notes are evidence-based. "Looking into it" is not a valid note. Include what you found, what you changed, or what's blocking you.

Response Format

All commands print JSON to stdout. The script handles auth headers automatically.

Valid Statuses

backlog, in_progress, review, done, cancelled