Menu
Determines minimal context and routes tasks to the correct prompts, agents and tools.
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
Based on SOC occupation classification
Guides the creation and review of Architectural Decision Records (ADRs). Use this skill when a significant architectural change is proposed, a new technology is introduced, or a trade-off needs to be documented.
Intelligently routes requests to the correct context instructions and documentation files. Use this skill to determine which AGENT.md, guide, or technical instruction file is most relevant for a user's request.
Scaffold API mocking infrastructure using Mountebank and Testcontainers for reliable integration testing.
Manage full-stack observability using Logfire (logging/tracing) and OpenObserve (storage/visualization).
Interface for managing the Temporal Database and AI system (ReDB + GGUF embeddings).
Spec-Driven Development workflow using VibesPro conventions
| name | context-routing |
| description | Determines minimal context and routes tasks to the correct prompts, agents and tools. |
| metadata | {"id":"ce.skill.context-routing","tags":["routing","context-min","context-debug"],"inputs":{"files":["ce.manifest.jsonc",".github/ce/routing-rules.md"],"concepts":["least-context"],"tools":[]},"outputs":{"artifacts":[],"files":[],"actions":["select-files","select-prompt","select-handoff"]},"dependsOn":{"artifacts":[],"files":[".github/ce/routing-rules.md"]},"related":{"artifacts":["ce.prompt.debug-routing"],"files":[]}} |
The context-routing skill applies deterministic rules to decide which files, prompts, skills and tools are necessary to answer a user request. Use this skill to minimise context loading and avoid hallucinations.
Load the manifest. Parse ce.manifest.jsonc to build an index of available artifacts,
their tags, inputs, outputs and dependencies. Validate that the manifest is well-formed.
Analyse signals. Inspect the user’s request to detect intent, scope, risk and
actionability signals (e.g. planning, implementation, review, debug). Use the routing
rules documented in .github/ce/routing-rules.md to map signals to candidate targets.
Select authoritative documents. Always include the core project documents
(PRODUCT.md, ARCHITECTURE.md, CONTRIBUTING.md) when relevant to the detected scope.
Use tags and dependencies to decide which docs are required.
Pick the primary prompt or skill. Based on the intent signal, select one prompt or
skill whose tags match the intent (e.g. planning → create-plan.prompt.md). Avoid
loading multiple prompts for a single request.
Resolve dependencies. For each selected artifact, load its dependsOn files and
artifacts. Ensure no more than the maximum configured in the manifest defaults
(maxFilesToLoad) are included unless explicitly requested by the user.
Return the route plan. Produce a list of target files, skills, prompts and tools to be loaded, along with a brief justification for each selection. If the plan includes any action that modifies files or executes commands, ensure that a validation task is also included.
Support debugging. When invoked via the debug-routing prompt, explain why each
artefact was selected and suggest any metadata or tag updates that would improve future
routing.
This skill makes routing decisions transparent and reproducible, enabling precise control over the AI’s context window and improving answer quality.
