I think a magnetic trap and a decently sized chunk of solid antimatter is the likeliest solution. This paper (pdf) briefly mentions the theoretical possibility of quantum reflection, making anti-atoms not "stick" to container walls. Very cool if it could be done, and maybe similar effects are used to keep metallic hydrogen compressed. This paper (paywall) suggests that storing antimatter as Einstein-Bose condensates might help. But both solutions likely require extreme cryogenics, everything has to be kept close to absolute zero. This paper (paywall) reviews different methods. It notes that for more than 10^12 positrons Penning traps require more energy than is stored in the antimatter. It also suggests that there might exist stable bound states where positrons sit inside normal molecules like C60 without interacting, and this might form very energetic materials. As I envision it, microscopic crystals of anti-atoms (likely anti-hydrogen or anti-lithium) with a magnetic moment are kept suspended within nanocontainers by very focused magnetic fields. These are likely at least partially generated by nanofactured super-strong permanent magnets, perhaps focused by small superconductor layers. The nanocontainers are rated for the accelerations they can take, probably in the tens of thousands of Gs. The system itself will be noticeable from occasional gamma photons and the presence of a sizeable magnetic moment (possibly shielded by superconducting plates on the outside). I don't think it is easy to get a shaped charge effect with a small antimatter pellet since momentum is mainly transferred in the form of hard gammas that tend to break nanocontainers: if the detonation starts at one side of a pellet it will break down containers across the pellet long before it imparts a directional motion by transferring momentum to the molecular lattice. However, a larger pen-like charge with gamma-absorbing layers might be able to act as a nuclear shaped charge: the initial detonation is absorbed by the first layer which expands, pushing the rest of the assembly forward, and then subsequent layers will continue this. Unfortunately (I'm bad at running nuclear detonation hydrodynamics code in my head) I think this will mainly pancake the assembly rather than make a nice beam of detonation.