obsfx

Start tracked work the right way: search first, restore task context, check dependencies, then mark the task in progress.

Close Trekker work cleanly by reviewing context, adding a summary comment, marking the task complete, and surfacing the next ready item.

Check Trekker for related tasks before creating a new one so work is extended instead of duplicated.

Suggest or manage Trekker tracking when the user talks about bugs, features, backlog items, or work that should survive this session.

Plan multi-step work in Trekker before implementation. Use epics, tasks, and dependencies so work survives session resets.

Search Trekker first to recover context before creating, starting, or updating tracked work.

Translate natural-language tracking questions into Trekker queries for status, history, blockers, and related work.

Work through Trekker tasks autonomously by recovering context, selecting ready work, documenting progress, and finishing with clean summaries.

Persistent task memory for AI agents across sessions

Use when user mentions issue, bug, task, ticket, feature, backlog, tracking, or work items. Suggests using Trekker for tracking.

Use Trekker to plan and track multi-step tasks BEFORE starting implementation.

Synchronize Trekker with Claude's built-in TodoWrite. Trekker is the source of truth.

Assisted learning skill for JavaScript developers exploring new technologies. Use when the user wants to learn, understand, or explore a new programming language (Elixir, Swift, Go, Rust, Python, etc.), a new tool or platform (Kubernetes, Docker, Terraform, Redis, Kafka, etc.), or a new architectural approach or paradigm (containerization, microservices, object pooling, distributed systems, event sourcing, CQRS, DDD, etc.). Triggers on: "teach me", "explain", "how does X work", "X for JavaScript developers", "learn X", "what is X", "compare X to JS", "I want to understand X", "show me how X works", "help me learn", "bridge from JS to X", "what's the equivalent of X in Y", or any question about unfamiliar technology from a JS developer's perspective. Does NOT generate project code — provides explanations, comparisons, diagrams, and insights. Stores learning progress in trekker as persistent memory across sessions.

Design, architect, plan, and review software structure. Use when the user says "design the architecture", "plan this feature", "how should we build this", "break this down", "propose approaches for", "what's the best way to structure", "create a technical plan", "think through the design", "evaluate options for building", "review the architecture", "analyze the codebase", "find pain points", "what should we improve", "audit the structure", or any medium-to-large task, structural change, refactoring, creative/idea work, multi-component change, architecture review, or technology selection. Produces multi-solution analysis with tradeoffs, composable module design, and architecture-first implementation plans.

Protocol for @feedback annotations in source code. Use when user mentions @feedback, address feedbacks, inline feedback, or when encountering @feedback markers during any task.

Use when Codex is working in a repo with `.trekker`, when the user asks to use Trekker, or when bugs, features, tasks, epics, dependencies, comments, handoffs, or backlog items need persistent tracking across sessions.