HardMagic / product
Field note · 2026.08 · agent skills
CLI Agent Skills
CLI Agent Skills connects agent skills, permissions, and evaluation as one consequential decision.
Position / 01
Small explicit skills outperform opaque all-purpose creative agents.
The decision in front of you
CLI Agent Skills connects agent skills, permissions, and evaluation as one consequential decision.
Small explicit skills outperform opaque all-purpose creative agents.
Why agent skills changes the work
This page helps readers connect HardMagic to agent harnesses.
It treats agent skills, permissions, and evaluation as connected operating concerns rather than isolated deliverables.
What to inspect
- Use the skill manifest to establish context, then inspect the invocation flow for mechanism and decision points.
- The permission envelope should expose evidence, trade-offs, or a practical exercise instead of serving as decoration.
Keep tracing the questionRead the skill contract ↗
Boundaries and proof
Capabilities, availability, integrations, and deployment details must reflect current product evidence and status.
Before publication, factual claims require inspectable product behavior, attributed first-party experience, or dated source notes.
Choose the next depth
- Ready readers can integrate a skill.
- Readers still framing the problem can read the skill contract.
Operating view
Make the invisible structure visible.
HardMagic works in systems: who decides, what evidence travels with the work, where intelligence runs, and how an idea survives the journey into production. This diagram is conceptual—not a claim of measured product performance.
The next useful depth
Choose the path that fits the decision.
Connect HardMagic to agent harnesses. If the question has an owner and a horizon, take the working path. If it still needs language, keep building context first.