AI Engineering

Execute Spec In Fork orchestrates one approved SPEC READY through a same-directory Codex App fork, runs spec-executor, validates the receipt, and archives the child when complete. Use it for explicit spec execution or to handle messenger replies and recovery events for that fork.

Goal Crafter turns vague tasks into verifiable goal prompts for AI coding agents like Claude Code, Codex Automations, and Pi. It is used when you need a self-running task with clear completion criteria and constraints.

Improve Codebase Architecture scans a codebase for shallow modules and deepening opportunities, then generates an HTML report of candidates. It is used to find refactors that improve locality, testability, and AI navigability before drilling into one choice.

Loop Me grills you about the workflows you want to build and turns the answers into workflow specs in this workspace. Use it when you want to define recurring loops, triggers, checkpoints, and briefs before implementation.

Spec Executor implements one approved spec in a forked execution thread and returns a structured receipt. Use it when a final SPEC READY block is approved and the work fits one implementation session.

To Spec turns the current conversation and codebase context into a project issue spec. It publishes the spec, applies the ready-for-agent label, and avoids interviewing the user.

Wayfinder plans large efforts as a shared map of decision tickets on an issue tracker, then resolves them one at a time until the route to the destination is clear. Use it when the work is too big for one agent session and needs decisions before execution.

Guides AI to structure responses for users with ADHD or high cognitive load, making tasks easier to start, resume, and finish.

Inspect captured HTTP traffic, WebSocket messages, and Sitemap entries using Caido MCP tools with compact projections, HTTPQL filters, and regex excerpts for safe, evidence-focused security workflows.

Improve an AI cover letter skill via controlled experiments with manuscript materials and expert benchmarks. Supports blind generation, route-specific training, and evaluation to avoid overfitting.

Conduct systematic source-backed research with scoped questions, triangulation, contradiction handling, and citation-ready synthesis.

Routes finance-related requests to the appropriate tools and capabilities available in the current Codex or Hermes runtime. Reports missing data, credentials, or libraries explicitly.