workbench-template

Config-driven visible plan-to-PR workflow for Workbench engineering work. Use for agent-flow, full lifecycle, visible agents, Helix-style implementation, cross-repo context, HTML plans, PR breakdowns, small or stacked PRs, reviewer packets, review from packet, Claude/Codex pairing, worktrees, Jira/PR handoffs, continuation handoffs, or approved bypass of detailed PR slicing.

Freeform ideation and structured shortlisting. Generates 8-10 approaches including wild cards, then converges on 2-4 viable options with tradeoffs. Use before /forge or standalone for any creative exploration.

Read Outlook Web calendar via Chrome and extract meeting data for capacity planning. Writes to context/active/calendar.md so other skills use real numbers.

Show remaining capacity today and this week by reading daily and weekly plans. Flags overcommitment. Use when checking if there's room for more work.

Read and critically summarize a Confluence document. Surfaces key decisions, gaps, open questions, and suggests review questions.

Create a release branch and managed _release worktree in a Workbench repo using canonical Python worktree tooling. Use when the user wants to cut a release, create a release branch, start a release process, preview the next release branch name, or keep release work out of main/develop.

Mid-day progress check against the daily plan. Scans for new DMs, checks capacity, and helps reprioritize if needed.

An interactive database design interviewer for relational databases (PostgreSQL, MySQL, SQL Server). Conducts a thorough, multi-phase interview to discover entities, user flows, access patterns, edge cases, and scale considerations — then produces ERD diagrams (Mermaid), access pattern analysis, and pros/cons of design tradeoffs. Use this skill whenever someone says "design a database", "help me think through a schema", "I need to model data for...", "database for a new feature", "walk me through database design", "interview me on my schema", or any variation where they want a guided, conversational approach to relational database design rather than just a schema dump. Also trigger when someone mentions "new database", "data model", "schema design", "ERD", "entity relationship", or "access patterns" in the context of building something new.

Decide whether to own a task or delegate it. Assesses principal-level requirement, complexity, and growth opportunity. Generates handoff briefs for delegation.

Review what happened today, capture unanswered DMs, update daily plan with completion data, and set up tomorrow. Run at the end of each work day.

Produce realistic time estimates for engineering tasks. Breaks down hidden work, applies calibrated multipliers, and gives three-point estimates. Counteracts chronic underestimation.

Unified Markdown-first feature planning pipeline. Chains brainstorm, intake, research, challenge, refinement, plan creation, external judging, and handoff. Replaces manual /plan-feature → /research → /sounding-board → /plan-create → /plan-resume orchestration. Generates HTML companions only when artifacts have substantial visual content (diagrams, mockups, semantic-color matrices).

Create managed hotfix worktrees and hotfix PRs using canonical Workbench worktree tooling. Use when the user asks for a hotfix branch/worktree, needs to cherry-pick or backport a merged PR to a release branch, needs a release-targeted hotfix PR, or asks for PPE/prod hotfix work while keeping develop/main clean.

Execute a forge plan step-by-step with test verification, progress tracking, and boundary checkpoints

Handle urgent interrupt requests. Sizes the impact, shows tradeoffs against current plan, and helps decide whether to take it on, delegate, or push back.

Pull a Jira ticket with linked docs, tickets, comments, and PRs. Produces an actionable summary with assessment and open questions.

Quickly log an interrupt to today's daily plan with a timestamp. Faster than the full interrupt agent for minor disruptions. Pass a brief description as the argument.

Gather context for an upcoming meeting from Jira, Confluence, and Slack. Produces a brief with background, open questions, and talking points.

Scan DMs, @mentions, Slack channels, and Jira to create a prioritized daily plan. Run this every morning to start the day focused.

Create a Markdown-first strategic feature plan with work breakdown and a handoff prompt for the target repo. Interviews for ambiguities, reads research/analysis artifacts if available. Use after research and refinement, or standalone when you already have context.

Gather full context for a PR — Jira ticket, design doc, Slack discussion. Produces a Markdown-first informed review brief so code review catches intent-vs-implementation gaps. Generates an HTML companion only when the brief has substantial visual content (architecture diagrams, side-by-side mockups).

Generate a Markdown-first Product Requirement Document (PRD) for a new or updated feature. Generates an HTML companion only when the PRD has substantial visual mockups, user-flow diagrams, or semantic-color comparison matrices. Works standalone or integrated into the forge pipeline. Use when creating PRDs, defining product requirements, scoping features, or when the user mentions PRD, product requirements, feature spec, or requirements document.

Test coverage and QA analysis with QA Lead posture — diff-aware gap detection, regression risk assessment

Track quarterly goals across teams. Set up new goals, add mid-quarter priorities, run weekly goal checks, and assess capacity impact of scope changes.

Multi-source research methodology with confidence ratings. Searches Confluence, Jira, Slack, and the web. Produces structured findings with source quality ratings. Use when investigating a technology choice, gathering context on a feature, or building an evidence base.

Adaptive code review orchestrator — classifies diff and invokes structural, security, QA, test, and CodeRabbit sub-skills

Security audit with Security Engineer posture — OWASP Top 10, secrets detection, dependency risks

Land implemented code — open the PR, monitor GitHub checks, merge, then clean up the worktree and local branch. Does NOT review (review-gauntlet is the authoritative review — run it first). Links Jira and updates plan state.

Create new skills, modify and improve existing skills, and measure skill performance. Use when users want to create a skill from scratch, edit, or optimize an existing skill, run evals to test a skill, benchmark skill performance with variance analysis, or optimize a skill's description for better triggering accuracy.

Scan Slack channels, DMs, and @mentions to surface what needs attention. Filters noise, prioritizes by urgency. Use after being away or to catch up mid-day.

Stress-test ideas with research-backed critical analysis. Steelmans positions, challenges assumptions, runs pre-mortems, explores alternatives. Use when making architecture decisions, evaluating technology choices, reviewing plans, or drafting communications.

Generate a quick 3-line standup summary (yesterday, today, blockers) from daily plan files. Use when the user needs a standup update or asks what they did yesterday.

Generate a stakeholder-ready status update organized by quarterly goals. Pulls from weekly plan, daily plans, and Jira. Markdown-first by default for agent / engineer readability; generates an HTML companion when the report is going to a non-technical stakeholder audience or has substantial visual content (status matrix with semantic colors, charts). Provides Slack/email paste formats when requested.

Structural code review with Staff Engineer posture — architecture, patterns, edge cases, maintainability

Scan worktree permissions and promote selected Claude Code permissions to global settings using canonical Python tooling. Use when the user wants to sync, promote, consolidate, or inspect permissions across Workbench worktrees.

Concrete test suggestions with TDD Coach posture — framework-aware skeletons, red-green-refactor guidance

Decompose a Jira epic or story into estimated subtasks including hidden work. Markdown-first saved artifacts (subtask lists are mostly tables/bullets); HTML companion only when the breakdown has substantial visual content (timeline SVG, dependency graph). Uses calibration data for realistic estimates. Optionally creates subtasks in Jira.

Create and maintain a realistic weekly capacity plan. Tracks actuals vs estimates, checks if new tasks fit, and manages the weekly cycle. Run Monday mornings.

End-of-week retrospective. Compares planned vs actual, updates calibration data, and surfaces estimation patterns. Run Friday afternoons.

Recommend the next highest-impact task based on remaining capacity, priority order, and pending DMs. Use when deciding what to work on next.