gateway

Produces a dispatch plan — isolation model, agent count, sequencing, and model assignments — for multi-agent parallel work. Use when: "dispatch agents in parallel", "run multiple tracks simultaneously", "plan parallel agent strategy", "should this be foreground or background".

Produces a structured background-dispatch handoff — task description, output contract, success criteria, and timeout — so long-running work completes without blocking the user. Use when: "run this in the background", "dispatch agent for long task", "I need this to run async", "parallel agent dispatch".

Routes health-audit findings to the correct execution path — main thread, foreground fix, or parallel agent waves. Use when health-audit findings need to be sorted by effort and dispatched as agent waves.

Produces a granular todo list where each item maps to one discrete deliverable or decision point, maintained at real-time accuracy throughout execution. Use when: 'what is the status', 'too many items in the list', 'simplify the progress tracking', 'long todo list'.

Conductor agent pattern that decomposes complex tasks, dispatches specialist sub-agents, manages dependencies, and synthesizes results into a unified deliverable. Use when: 'coordinate multiple specialist agents', 'orchestrate a complex multi-phase task', 'run a conductor pattern across agents', 'this task needs live agent coordination', 'decompose and dispatch to sub-agents'.

Selects the correct multi-agent orchestration pattern for a task using an 11-signal selection matrix. Use when choosing which orchestration pattern fits a multi-agent task.

Produces a verified multi-track execution by dispatching parallel agents with file manifests, independent verification, and status tracking. Use when: "dispatch agents", "parallel tracks", "multi-track work", "fan out", "divide and conquer".

Produces a structured codebase map — directory roles, entry points, dependency graph, architectural patterns, and dragon locations — for any unfamiliar repository. Use when: 'map this codebase for me', 'I need to understand this repo before I start', 'give me a reading order for this project', 'where are the dragons in this code'.

Produces a scored comparison of available DESIGN.md design systems against a project's requirements and delivers a recommended selection with adaptation notes. Use when: 'which design system fits this project', 'pick a design reference for me', 'match a design system to this brand', 'compare design system options for this client'.

Extracts design tokens, component structure, and layout from a Figma file and produces framework-specific code (HTML, Svelte, React, or Vue) that preserves design intent rather than pixel-copying. Use when: 'implement this Figma design', 'convert this Figma file to code', 'extract tokens from Figma', 'build this component from the design'.

5-layer code analysis (AST, Call Graph, CFG, DFG, PDG) that yields 95% token savings over raw file reads. Use when: 'systematically understand an unfamiliar codebase', 'analyze code structure with minimal tokens', 'map call graphs and data flow', 'find dead code and architectural layers', 'trace variable definitions across files'.

Produces a typeset PDF artifact from any source file — code, Markdown, HTML, chat JSON, or structured Dojo files (specs, ADRs, skills). Use when: "export this to PDF", "generate a PDF of this file", "make a PDF from this".

Configure Pretext typography settings for PDF export — font stacks, sizing, line height, page geometry, and disposition-driven typography mapping. Use when defaults need tuning for specific content types or audiences.

Produces a signed distribution manifest and CAS-packaged skill set by orchestrating the full community-to-Dojo pipeline: shallow-clone repos, scan for compatibility, normalize frontmatter in parallel batches of 10, package via dojo CLI, and emit manifest.json with source attribution for every entry. Use when: "import community skills", "batch normalize skills", "package skills from repos", "onboard external skills", "run the skill supply chain".

Produces an enriched SKILL.md with all six Dojo SkillRegistry.IsValid() fields populated — name, description, tier, agents, tool_dependencies, and trigger phrases — inferred from the existing body without rewriting it. Use when: "normalize this skill", "make this skill dojo-compatible", "import a community skill", "enrich skill frontmatter", "fix skill registry validation", "prepare skill for dojo", "community skill is missing fields".

Produces a compatibility catalog — skill-scan-report.md and skill-scan-catalog.json — classifying every SKILL.md in one or more GitHub repos or local paths as ready, normalizable, or incompatible, so you know exactly what normalization work lies ahead before committing to an import. Use when: "scan this repo for skills", "check which skills are dojo-compatible", "audit community skills", "catalog skills from a github repo", "find skill files in this repository", "assess community skill compatibility", "build a skill import report".

Produces a quantified current-state report (test count, aria markers, error boundaries, storage usage, dependencies) used to anchor specifications in measured reality rather than assumptions. Use when: 'audit the codebase before speccing', 'ground this spec in reality', 'what does the code actually look like', 'measure before I write the spec'.

Spec-to-fix pipeline: audit a codebase against its specification, classify gaps by severity, fix in parallel, track in GAPS.md registry. Use when a spec exists but implementation completeness is unknown.

Produces a gap inventory and Track 0 remediation commit that closes mismatches between implementation prompts and actual codebase state, ensuring parallel tracks start from a clean, type-safe foundation. Use when: 'commission these tracks', 'run Track 0 before handing off', 'check for spec-to-code drift', 'align before we build'.

Adversarial code review that breaks the self-review monoculture. Use when you want a genuinely critical review of recent changes, before merging a PR, or when you suspect Claude is being too agreeable about code quality. Forces perspective shifts through hostile reviewer personas that catch blind spots the author's mental model shares with the reviewer.

Produces an aggregated markdown performance report from stored trace data — tool call metrics, token usage, cost, bottlenecks, and numbered recommendations. Use when: "session cost report", "what ran this week", "why did this spike", "end-of-sprint performance review", "budget alert fired".

Produces a structured APPROVE / WARN / BLOCK decision by checking remaining token budget across query, session, and monthly tiers before an expensive operation runs. Use when: "is this operation within budget", "pre-flight check before web search", "budget alert fired", "before a multi-step pipeline", "session above 70% utilization".

Produces a verified build/test status report and auto-fixed modules by sweeping all Go modules in a workspace. Use when: "run a sweep", "check all builds", "nightly sweep", "fix all failing modules", "workspace health check".

Produces a CLAUDE.md health report listing conflicts, redundancies, and stale rules across the global, project, and subdirectory hierarchy — and optionally installs a PreToolUse hook to block unauthorized modifications. Use when: "agent behavior is inconsistent", "just merged branches that touched CLAUDE.md", "onboarding a new agent", "after applying learnings", "weekly maintenance".

Runs a structured 7-phase convergence session to arrest strategic drift after 3-4 Sonnet feature sessions. Use when the drift-detector fires YELLOW or RED, or when you notice accumulating uncommitted state, deferred validations, and growing open-items lists.

Produces a hook script scaffold and settings.json registration block by looking up the correct lifecycle event, exit code pattern, and configuration for a Claude Code hook. Use when: "write a new hook", "hook is not firing", "block a tool call mechanically", "configure sub-agent coordination", "implement a security block pattern".

Scaffolds new MCP servers with tool definitions, transport wiring, handler structure, and test harness for Python (FastMCP) or TypeScript (MCP SDK). Use when: 'build a new MCP server', 'scaffold an MCP integration', 'create tools for an external API', 'set up MCP transport and handlers'.

Produces a running multi-agent observability dashboard by deploying the hook capture pipeline, SQLite WAL store, and WebSocket-streaming Vue client — then validates it with a live test session. Use when: "set up monitoring for multi-agent workflows", "debug parallel agent coordination", "audit tool call history across a session", "build visibility for a demo or review".

Produces an architectural specification document mapping Gateway SSE events and OTEL spans to named dashboard widgets, a data flow diagram, and an MCP App config — ready to hand to an implementation agent. Use when: "design the observability dashboard", "spec the monitoring app", "map events to widgets", "plan the agent dashboard", "extend the dashboard with new widget types".

Produces a provenance determination — pointer or gap — for empty or apparently incomplete directories, with a documented rationale and recommended action. Use when: "this directory is empty", "missing SKILL.md", "coverage metric shows a gap", "why is this directory here with nothing in it", "before creating content to fill an apparent void".

Produces a Supply Chain Refresh Report and updated manifest.json by pulling the latest from all watched repos, scanning for new or changed skills, normalizing, and packaging them into CAS. Use when: "refresh the supply chain", "update community skills", "weekly skill library maintenance", "a community repo announced a new release".

Produces structured OTEL-compatible log entries and a session summary (call counts, success rate, cost, top tools by cost and frequency) for all tool executions within a session. Use when: "log tool calls for this session", "debug unexpected tool selection", "audit token cost by tool", "validate a new MCP integration", "build a session summary for handoff".

Map an external AI tool's behavioral patterns to Dojo ADA disposition fields by analyzing its ingested system prompt. Produces a disposition YAML approximation and similarity score against Dojo defaults. Use when understanding how other agents think or when designing new disposition presets. Trigger phrases: "analyze this agent's behavior", "map prompt to disposition", "compare agent behavior", "what disposition does this tool use", "reverse-engineer agent personality".

Produces a cross-tool disposition matrix, behavioral cluster analysis, and a set of named YAML disposition presets by synthesizing analyses from 3+ ingested AI system prompts. Use when: "build an intelligence map", "compare all agent behaviors", "generate disposition presets from external tools", "what can we learn from other tools", "synthesize agent analyses".

Maintains a running decision log and context state across agent sessions, preventing context loss between clears and session boundaries. Use when: 'starting a new session on an ongoing project', 'context is getting heavy', 'preserve state before clearing', 'create a handoff document', 'pick up where I left off'.

Produces a stored MemoryEntry (structured sections + ADA disposition indicators) and one or more MemorySeeds by parsing a system prompt from an external AI tool into Dojo memory. Use when: "ingest this system prompt", "store this prompt in memory", "parse this agent's rules", "import this Cursor prompt", "analyze what this prompt does".

Captures user corrections from a session, validates them semantically, and writes approved learnings to the most specific persistence target (skill file, project CLAUDE.md, or global CLAUDE.md). Use when: "remember this for next time", "don't do that again", "update your behavior", "capture this correction", "reflect on what we learned".

Produces updated memory artifacts by compressing a session's key decisions, changes, and context into the structured memory garden. Use when: "compress this session", "save context", "end of session", "update memory", "wrap up".

Designs and configures all three loops of solo-operator session automation — session bookends, background maintenance agents, and a delegation flywheel. Use when setting up a new workspace, recovering from the 44% zero-artifact session leak, or preparing to scale delegation across multiple concurrent agents.

Produces a named ADA disposition preset (YAML) and a layered behavioral analysis (structural decomposition + evidence-cited classifications) from a system prompt. Use when: "analyze this system prompt", "extract disposition from prompt", "reverse-engineer agent behavior", "what patterns does this prompt use", "build a preset from this prompt".