Scores tasks on five axes to recommend capability bands and effort levels for model routing.
npx skills add https://github.com/allemaar/open-skills --skill route-task-allemaar
Please help me install this Agent Skill. Skill: route-task Source: https://github.com/allemaar/open-skills/tree/main/skills/route-task Command: npx skills add https://github.com/allemaar/open-skills --skill route-task-allemaar
High-intent search queries and answers about installing and using this skill.
Score the task on five axes (Risk, Uncertainty, Scope, Coupling, Verification) using the rubric's anchored 0-4 scale. The maximum of Risk and Uncertainty maps to a capability band from 0 (cheapest capable) to 3 (strongest at maximum effort), which the runtime resolves against its own live model roster.
Rate each of the five axes from 0 to 4 using the anchor definitions, then compute capability as max(Risk, Uncertainty) mapped to 0-3 and volume as (Scope + Coupling + Verification) times 2. Emit the full route record with one evidence sentence per axis as the audit trail.
No, the rubric only recommends and never mandates. It never overrides an explicit handler model choice, never authorizes spend or dispatch, and degrades to the session model whenever the roster is ambiguous or a ladder position cannot be resolved.
The 1 and 3 anchor scores are deliberately undefined pending calibration, and the max(Risk, Uncertainty) formula cannot distinguish combined high scores from a single high axis. Coupling's dual role as floor and volume input is unvalidated, so surprises should be logged in the report card's feedback field.
The plan gate fires as design_open when design or architecture is genuinely unresolved, an architecture finding demands it, or the handler requires a plan. It is evidence-derived and independent of the capability band, signaling that a phased plan should precede the build.
radius-projectGenerates code-grounded architecture documentation and Mermaid diagrams for Go codebases.
cuiossManage plan and per-phase Q-Gate findings via JSONL CLI workflows.
Snausage0x45Analyze binary executables for malicious intent via static and dynamic emulation.
qodex-aiRun verification commands and present output before claiming task completion.
