Auto-detect task complexity for Claude models (Haiku + Sonnet). Route simple tasks to Haiku, escalate complex ones to Sonnet. Save 50%+ on API costs. | 適用 Claude 模型（Haiku + Sonnet），自動偵測任務難度，簡單用 Haiku，複雜升級 Sonnet，輕鬆省下 50% 以上的費用。— RULES: Before responding, classify the task. If it matches an escalation trigger, call sessions_spawn with the full task and model anthropic/claude-sonnet-4-20250514 IMMEDIATELY. Do NOT answer complex tasks yourself. ESCALATE: analysis, comparisons, planning, strategy, c
Security Analysis
high confidenceThe skill's instructions and requirements are internally consistent with its stated purpose (routing tasks between Haiku and Sonnet), but it will routinely forward user task text to the higher-cost Sonnet model when its escalation rules match — so consider privacy and cost implications before enabling.
Name/description promise (route simple tasks to Haiku, escalate complex ones to Sonnet) matches the SKILL.md rules which explicitly classify tasks and call sessions_spawn to spawn the Sonnet model when escalation triggers are hit. No unrelated binaries, env vars, or config paths are requested.
SKILL.md narrowly confines behaviour to classification and conditional escalation. However, the instructions mandate forwarding the full task to Sonnet for any matched trigger (and say 'when in doubt, escalate'), which grants the skill broad discretion to transmit user-provided content to another model. That forwarding is coherent with the purpose but increases privacy exposure: sensitive input will be sent to the spawned session whenever triggers apply.
Instruction-only skill with no install spec and no code files — nothing is written to disk and no external packages or download URLs are used. This is the lowest-risk install profile.
The skill requests no environment variables, credentials, or config paths. There are no disproportionate or unexplained secret requests.
always is false and the skill does not request persistent or system-wide configuration changes. It instructs use of sessions_spawn (normal platform capability) but does not attempt to modify other skills or global settings.
Guidance
This skill appears to do what it says (route simple tasks to a cheaper model and escalate complex ones), and it has no install or secret requirements. Before enabling it, consider: 1) privacy — the skill will forward the full task text to the Sonnet session whenever escalation triggers match (and the rules encourage 'when in doubt, escalate'), so avoid sending highly sensitive data through this flow unless you're comfortable with how spawned sessions are logged and where they run; 2) cost — frequent conservative escalation will increase API usage/costs (the skill explicitly favors escalation over risk of a bad Haiku answer); 3) test behavior — try with non-sensitive queries to observe how often it escalates; and 4) review platform/session_spawn policies — confirm where Sonnet runs, retention/logging, and whether you can restrict or audit spawned sessions. If you need stricter privacy/cost controls, ask for a version that exposes fewer 'when in doubt' cases or that requires explicit user confirmation before escalation.
Latest Release
v4.2.0
Add Claude model mention, complete 費用 text, add body explanation for long description
More by @peterokase42
Published by @peterokase42 on ClawHub
