AI Agent Workflow Collection

Connect to an Android phone over ADB wireless (Galaxy S25 at meain's setup), install/launch APKs, take screenshots, and drive the UI via uiautomator. Triggers: "connect to phone", "adb connect", "test on phone", "screenshot the phone", "install on phone", "phone is at 192.168.1.3", any task that needs to drive an Android device from the host.

Look up a URL (Teams message, Jira ticket, GitHub PR/issue, incident.io alert) and create a concise backlog entry with a [ref] link.

Update the daily backlog by gathering data from Jira sprint, GitHub PRs, and existing backlog, then collaboratively deciding what to work on today. Handles checked-off items, moves existing tasks to Today, and adds new items with reference links. Commits the backlog folder before making changes.

Query and analyze codebases semantically using codebase-memory-mcp graph database. Use to explore code structure, complexity, dependencies, call graphs, and architecture.

Generate an interactive local HTML review page for a large PR or diff, grouping the changed files by logical concern (not just by file) so a reviewer can go through one theme at a time instead of a flat file list. Each concern gets its own section with a description, per-file summaries, the real diff (sliced to just the hunks relevant to that concern when a file serves more than one), and callouts for anything genuinely non-obvious. Triggers: /concern-review, "review this PR by concern", "review this large PR", "build a review dashboard", "group this diff by concern", "concern review", "help me review this PR".

Copy text to clipboard as rich HTML so it pastes with formatting into Microsoft Teams.

Verify a claim, assumption, or answer before acting on it or reporting it to the user. Use when the user says /double-check, "double check that", "verify that", "are you sure", "check again", or when you've just made a factual claim that could be verified (e.g. "no reviewers added", "the file doesn't exist", "the test passed") and the user questions it.

Open a draft in emacsclient (blocking) so the user can review or edit it before further action.

Query Elasticsearch/Kibana using ES|QL via the Kibana async search API. Requires an initial curl command from the user to extract session credentials.

Explain how code flows with concrete input/output examples, ASCII diagrams, and before/after tables. Works on PRs, functions, modules, or any code path.

Control Firefox browser on macOS: open URLs, close/switch tabs, get active tab info, focus Firefox, send keystrokes. Use when asked to open a link in Firefox, close a tab, switch to a tab, find what's open in Firefox, or automate any browser action.

Respond as Michel — draft messages, replies, or any written content in the user's voice and personality.

Generate a concise weekly status update in team format by pulling data from the current Jira sprint (tickets assigned to me), backlog, and recent activity. Outputs bullet points covering "What I worked on" (completed/in-progress this sprint), "What's next" (To Do in current sprint), and standard blockers/leaves fields. When asked for last week's report, covers completed items from last week. Upcoming items are based on the current sprint's To Do tickets.

Lightweight code review via subagent. The main agent identifies commits and passes them to a subagent that retrieves the changes and reviews for correctness issues, consistency problems, and gaps. No praise, no style nits. Use when the user says /quick-review, "review my changes", "review this diff", "sanity check my code", or after finishing a feature/fix and wanting a second opinion before pushing.

Review code for refactorability — surface concrete, prioritized refactoring opportunities grounded in Martin Fowler's smell catalog and SOLID, augmented with static analysis tools (gocyclo, gocognit, dupl, etc.) when available. Go-first, polyglot-aware. Defaults to the current diff. Produces a High/Med/Low report with specific refactoring moves and a "Do NOT refactor" section to prevent over-eager suggestions.

Save an AI-generated artifact to the artifacts repository with structured metadata.

Summarize and save research findings from the current conversation into a well-structured markdown document.

Manage tasks using Taskwarrior (task CLI) - create, list, update, complete, annotate, and manage dependencies between tasks.

Run a command in a tmux pane/window/session and monitor its output for errors.

Manage the Obsidian vault at ~/.local/share/sbdb -- backlog tasks, meeting notes, journal entries, task notes, and general vault lookups.

Control VibeDiff code review instances — fetch review comments, reply to threads, create comments, resolve/reopen threads, and manage review state. Use when the user says /vibediff, "check vibediff", "get review comments", "reply to review comment", "resolve vibediff comment", or when working on code changes that have active VibeDiff review comments.

Explain a jj revision by fetching its diff and adding explanatory comments in VibeDiff to document WHY changes were made. Use when the user says /vibediff-explain, "explain this commit in vibediff", "add comments to vibediff", "explain this revision in vibediff", or asks to annotate a diff with explanatory comments.