ObneyAI/orc / orc-skills/orc-design-tree
orc-design-tree
VerifiedDesign a hand-assembled ORC behavior tree from a knowledge-work goal, including known specialist selection and revisiting, typed context, success conditions, control nodes and bounded recovery.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 15.3 KB |
| references/agent-design-literature.md | markdown | 3.9 KB |
| references/authored-goal-solving.md | markdown | 19.8 KB |
| references/bounded-repair.md | markdown | 11.0 KB |
| references/bt-assembly.md | markdown | 12.6 KB |
| references/evidence-closure.md | markdown | 2.8 KB |
| references/first-use-contract-check.md | markdown | 3.3 KB |
| references/specialist-design.md | markdown | 15.1 KB |
| references/specialist-discovery.md | markdown | 3.4 KB |
references/evidence-closure.md
View on GitHub# Complete the investigation before claiming the goal
Use this when an authored agent can consult tools or specialists and decide whether the task is complete. A truthful summary of the currently visible evidence can still stop too early: an answerable material question may remain open.
## Distinguish evidence states
Represent not-yet-requested, retrieved, observed-unavailable, superseded and irrelevant-to-this-goal separately where they affect decisions. An empty field only says the value is absent from this packet. It does not prove the world lacks the evidence, that the tool cannot obtain it, or that another specialist already checked it.
Give the completion checker the original goal, current artifacts, material qualifications, and relevant evidence-access options. The routing model also needs these facts. Putting a tool catalog only on the router leaves a preceding critic unable to tell whether a gap is answerable. A router can then accept an authoritative-sounding but underinformed critique.
A useful evidence question records its obligation, why the answer matters, who owns it, the available described resolver, observation state, and the actual answer or blocker. Keep this focused: unrelated tools and the entire execution history do not belong in every model call. The source inventory describes what can be inspected; it does not reveal or invent unread contents.
## Close questions semantically
Ask whether a remaining question could change the requested conclusion or action. Consult a suitable available source when it could. Re-enter the original owner with the answer, relevant prior artifact and current feedback, then reassess the parent goal. Ownership transfer is a different action and needs its own explicit contract.
Do not require every tool to run. A current authoritative prohibition can make an older optional exception irrelevant. A fully supported negative recommendation can satisfy decision support. Conversely, calling an unresolved task complete merely because the wording says “on current evidence” changes the goal without permission. Preserve an explicit wait or handoff when a material question really cannot be answered.
## Prove both stopping and continuation
Use one deficient artifact with an answerable material gap, and one correct qualified artifact that should be accepted. Verify actual resolver invocation, answer transfer, original-owner resumption and retained independent work. Inspect exact model inputs, not only declared reads. Check that the same completion contract permits the valid negative outcome without demanding irrelevant work.
A model critic is evidence, not an oracle. A finite decision model can select the next route; generated feedback helps when revision needs more than a choice. Neither a schema-valid result nor a critic's PASS establishes that all material questions were answered.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| brainstormingobra/superpowersYou MUST use this before any creative work - creating features, building components, adding functionality, modifying behavior, or planning anything new, in software or out of it (a talk, a business, a renovation). | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| brand-guidelinesanthropics/skillsApplies Anthropic's official brand colors and typography to any sort of artifact that may benefit from having Anthropic's look-and-feel. Use it when brand colors or style guidelines, visual formatting, or company design standards apply. | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |
| frontend-designanthropics/skillsGuidance for distinctive, intentional visual design when building new UI or reshaping an existing one. Helps with aesthetic direction, typography, and making choices that don't read as templated defaults. | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |
| mcp-builderanthropics/skillsGuide for creating high-quality MCP (Model Context Protocol) servers that enable LLMs to interact with external services through well-designed tools. Use when building MCP servers to integrate external APIs or services, whether in Python (FastMCP) or Node/TypeScript (MCP SDK). | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |