All 94 skillsPlan and decide15 of 15
stryx:orchestrate
When a job could use more than one agent: how to split it, staff it and merge the results.
/stryx:orchestrateInside its piece: 10 specks, one for each file on its shelf. Its fourteen group-mates glow on the open book behind it. The light splits toward red, the band of Plan and decide.
In numbers
counted 2026-10-12 in the installed set- 8reference files
- 10files in the skill
- 148KB on disk
- 18chapters
- 2agents work by this book
Use it when
in the skill’s own wordsWhen work could use more than one agent
- splitting a task into parallel subagents
- forming an agent team
- racing 2-4 competing approaches when the right fix or implementation is uncertain
- running a dynamic workflow
- deciding to stay solo
- Also when briefing or staffing subagents (which agent type, model, effort, tools), monitoring running agents, or merging their results
- Triggers
- "use agents"
- "parallelize"
- "fan out"
- "speed this up"
- "build a team"
- "spawn an agent"
- "delegate"
- "orchestrate"
- "try different approaches"
- "race approaches"
- "ultracode"
- "which model for this agent"
- "what are my agents doing"
- multi-file or multi-domain changes
- bugs with several plausible causes
- a subagent's report that would flood the context
Not for
another skill does it- Choosing between non-executable design options→ stryx:bakeoff
- Executing a written plan task by task→ stryx:execute
- An agent inside an app→ stryx:ai-agents
- Designing a new agent and writing its file→ stryx:skill-forge
Inside the book
18 chapters, by title only- The rule: findings go to files, only status comes back
- Procedure in short
- Step 1 - Pick the mode
- Step 2 - Decompose
- Step 3 - Staff
- Step 4 - Brief
- Context Project, stack, conventions, why this task exists. Absolute project root.
- Your task One specific, concrete deliverable for THIS agent, plus the absolute paths it must READ
- Files you own Absolute paths this agent may create or modify (a read-only agent: its findings file only).
- Do NOT modify Files owned by other agents (or "everything else").
- Constraints Patterns to follow (point at existing code), what not to do, no scope creep.
- Done when How to verify: e.g. npx tsc --noEmit shows 0 errors, named tests pass.
- Return Write findings to <run folder>/NN-<name>.md, written out as a full absolute path
- Step 5 - Launch
- Step 6 - Monitor
- Step 7 - Collect, synthesize, verify, report
- Detail tables from the steps
- Success criteria
Its helpers
2 agents work by this book- stryx:handoff-specialistWhen a session is ending with work left: it writes the handoff file and checks it.
- stryx:orchestratorWhen a large job has many parts: it splits the work, sends specialists and checks their results.
Works with
the skills it names most- stryx:bakeoff→When two or more very different approaches must each be drafted before you can choose.
- stryx:execute→When a written plan must be carried out to the end, or you say "keep going until it passes".
- stryx:skill-forge→When creating or changing a skill, agent, slash command or hook, or a skill did not fire.
- stryx:ai-agents→When building or fixing an AI assistant inside your own app.
- stryx:verify→When Claude is about to say done, fixed or ready to commit: fresh proof comes first.
- stryx:ai-power→When you are not sure which skill fits, or you say "load aipower": it is the map of all skills.
Claude loads it by itself when the moment comes; by hand, type the line at the top. It comes with the whole set: Installs all 94 skills, 32 agents and the brain program as one set. The skill’s own text is private source and is not shown here.