WHEN HOTLINE
Choose Hotline when...
Use Hotline when you want a callable capability with private execution logic, usage signals, and a seller-side business model.
Use Hotline
KEY TAKEAWAYS
In three lines
- Skills expand prompt context; Hotline moves work outside prompt context.
- Skills expose instructions; Hotline hides implementation behind a contract.
- Skills are great for behavior shaping; Hotline is better for monetizable execution.
DIMENSIONS
Dimension by dimension
Each row compares what Hotline standardizes, what Agent Skills focuses on, and why the difference matters.
Dimension
Hotline
Agent Skills
Execution location
Runs at the Responder side.
Why it matters
Execution location decides whether private logic stays private.
Runs inside model-guided prompt execution.
Prompt footprint
Near-zero prompt footprint beyond discovery metadata.
Why it matters
Large skill stacks compete directly with task context.
Consumes prompt space and token budget.
Commercial model
Supports per-call packaging and settlement.
Why it matters
Seller-side economics need to be explicit if the capability is the product.
Usually free or monetized outside the skill itself.
Source visibility
Implementation remains black-boxed.
Why it matters
If your moat is in workflow design, prompt exposure matters.
Instruction logic is usually visible to the model.
WHEN AGENT SKILLS
Choose Agent Skills when...
Use Skills when your main goal is to steer model behavior inside a single agent environment and you are comfortable exposing that guidance.
Not necessarily Hotline
PAGE FAQ
Common questions about Agent Skills
Can a Skill point to a Hotline?
Yes. That is a strong composition pattern: use a Skill to teach the model when to call a Hotline, then let the Hotline perform the protected execution.
Does Hotline replace every use of Skills?
No. Skills are still useful for instruction, orchestration hints, and local behavior shaping. Hotline becomes valuable when the capability itself is the product or needs runtime isolation.
Why is prompt isolation so important?
Because proprietary workflows, prompts, or domain heuristics lose value if they must be injected into the caller's context every time the agent needs the capability.