David Ondrej

Prepare read-only PostgreSQL access for AI agents, including SQL files, protected connection settings, and checks that writes and sensitive-table reads fail. A human applies production database changes and agrees which tables agents must not access.

• Prepare a read-only database account for agents
• Exclude tables containing secrets or personal data
• Verify that writes and restricted reads are blocked

Create, install, and debug TypeScript plugins that extend bb with custom pages, sidebar panels, CLI commands, agent tools, themes, and other features. The skill also guides development reloads, log checks, and verification of installed plugins.

• Add a custom page or sidebar panel to bb
• Create a bb CLI command or agent tool
• Investigate plugin errors through logs

Turn a research question into a focused DeepAPI request and save the findings as a Markdown report with source links. For longer reports, research subquestions separately and combine their answers and sources into one file.

• Research a question that informs a decision
• Save findings with citation URLs
• Combine separate subquestions into a longer report

Automate web tasks in an already-running Chrome session, including clicks, forms, and content extraction from signed-in pages. Use screenshots to guide actions and verify visible results, with remote browsers available for isolated work.

• Extract content from pages that require sign-in
• Automate clicks and form interactions
• Check visible results with screenshots

Set up and manage automatic Git syncing for shared documents, notes, and skills on macOS. The skill guides repository registration, removal, service checks, and maintenance, with local changes automatically committed and pushed by repo-sync.

• Setting up automatic syncing for shared team notes
• Adding or removing a repository from automatic syncing
• Checking sync failures and maintaining the background service

Configure a shared guard that blocks destructive shell commands before local AI agents execute them. Adjust blocking rules, connect agent integrations, and test why commands are blocked or allowed.

• Adjust rules for blocking destructive commands
• Connect the guard to another agent or machine
• Investigate blocked or unexpectedly allowed commands

Assess online purchases and subscriptions by comparing prices, checking seller trust, and considering delivery. It summarizes suitable places to buy and relevant business tax checks, while leaving orders, accounts, and payment details to the user.

• Checking whether a listed price is fair
• Comparing trustworthy shops and delivery options
• Investigating an unfamiliar seller

Create and manage Boat cloud sandboxes, remote machines for development work, through API, CLI, or SSH. Configure repositories and secrets, prepare reusable templates, manage machine state, and verify coding agents inside the selected environment.

• Create a cloud workspace for a coding agent
• Configure repositories and secrets
• Reuse prepared environments through templates

Keep a running list of unresolved topics and decisions during a conversation. The agent works through one topic at a time, updates the remaining items after each response, and removes items once resolved.

• Tracking unresolved decisions during a discussion
• Reviewing several topics one at a time
• Checking what remains before moving on

Request an independent code review using GPT-6.1-Sol at Max reasoning through Codex and Cloudroom. Review the current environment, including uncommitted changes, and return the completed reviewer’s full response covering serious issues and readiness to merge.

• Get an independent review before merging
• Review changes that are still uncommitted
• Check serious issues across related code and tests

Handle Git and GitHub CLI commands blocked by a sandbox, an environment that restricts access. Retry only authorized commands on the host, then check authentication, repository status, or pull request details there.

• Checking authentication after a sandbox failure
• Handling blocked GitHub repository and pull request commands
• Retrying Git writes blocked by permission errors

Review important choices the agent made during the current work and identify those it remains unsure about. It considers overlooked alternatives and returns a concise list in plain English when explicitly requested by name.

• Reviewing uncertain choices during ongoing work
• Checking for alternatives that were overlooked
• Finding decisions that need further discussion

Manage local development servers and APIs with macOS LaunchAgents, which run background processes. Keep servers running after the shell closes, restart them after crashes, inspect logs, and handle occupied ports when starting a service.

• Keeping a local server running after closing the shell
• Restarting a server automatically after a crash
• Checking server status and logs