There are a lot of reasons for the non-awakened to be leery of magic. Perhaps the biggest one is that spells like stunbolt and manabolt are not obvious. A perceptive runner will notice when a potential adversary aims a firearm or baton at you, but your first indication of offensive magic may be when you start bleeding out of your ears.

Flipping the roles around, what this means for the spellcasting runner is that in cases where adversaries are not aware that they are in fact adversaries, casting a stunbolt or manabolt at deadly means they may not realize they're under attack until after you've knocked one of them out.
NOTE: Spellcasting alone does not initiate combat, meaning it does not put you in initiative rounds. Outside of initiative, you have a half second delay (IRL time) between spells.
NOTE: If you deal damage, you enter combat and so your character will respond accordingly (drawing a weapon if you have a readied sheath/holster and attacking automatically if you have no other actions/spells queued up). Some spells/actions will cause combat without dealing damage (e.g., ignite, banishing). If you want to change this behavior, you can explore TOGGLE AUTOKILL and TOGGLE PASSIVE.
NOTE: Mobs can become alarmed if they notice your spellcasting. A higher magic rating and a lower spell force makes the spellcasting harder to notice, but a lower spell force is also easier to resist.
NOTE: In combat, spell casting is queued. Each time you send a CAST command, your spell is added to the queue. On your action in the combat loop, if there's at least one spell in your queue, you pop the next spell off the queue. If your spell queue is empty, you auto-attack with whatever weapon you have in hand (or otherwise with your fists) instead of casting.

While manabolt and powerbolt deal physical damage, stunbolt deals mental. This means that using stunbolts, a spellcaster needs to deal twice as many boxes to mortally wound a target, first filling their mental track before overflowing into their physical track. However, stunbolt has the advantage of lower drain. Spell drain codes are calculated as (F/2 + modifier)(drain level), i.e., casting a stunbolt 6 at deadly has a drain of (6/2 - 1)(deadly) = 2D. Let's take a look at the odds to completely resist the drain for stunbolt, manabolt, and powerbolt, all cast at force 6 and deadly, compared to drain resist dice (willpower + spell pool). A trauma damper does make this significantly easier.

Another attack spell that is often recommended is waterbolt. While stunbolt/manabolt/powerbolt are direct combat spells, waterbolt acts more like a firearm - it can be dodged and armor applies (though at half impact instead of full ballistic). But unlike stunbolt, it can be used against non-living opponents like turrets and you only roll against TN 4 (plus vision, wound, and sustain modifiers) regardless of how high your target's attributes may be.

Improved invisibility is probably the most desired (or hated, depending on who has it) spell in the game. But for runners, there is a bit of nuance to using it. If the invisibility is seen through by any means, guards will respond aggressively. While a force 1 stealth spell can be used to prevent auto-detection via ultrasound, astral perception or the rare lucky spell resistance roll will still be able to spot you. In addition, if you initiate combat in a way that does not immediately knock your target unconscious, they will fight back even if they can't see you. Once somebody is fighting you, they can also alert others who cannot see you and so you can end up fighting everybody in the room (and possibly also adjacent rooms). There are a number of autoruns that will go smoother if the subtle runner does not use invisibility.
For the not-so-subtle runner, improved invisibility is a very potent buff, as adversaries who are fighting you but can't see you suffer a +8 TN blind fire penalty. If they have ultrasound, they suffer a penalty of (8 + the force of your stealth spell)/2 (round up), max +8. That said, it's not a stand-alone solution; if you don't have the dodge or melee dice, then they can still luck into landing the occassional hit, and without soak and armor, a hit with just a single net success can still do a lot of damage.

The heal and treat spells can recover one box of physical health per success, to a maximum equal to the force of the spell. But the number of successes we can expect is highly dependent on the TN, which is set by 10 - (target's essence, rounded down) + TN penalties (e.g., sustaining, wounds, vision). Rolling 12 dice (e.g., 6 sorcery + 6 spell pool) against TN 4 (i.e., zero cyberware and no other penalties) averages 6 successes (thus 6 boxes healed assuming force 6+), whereas against TN 12 (say, 2 essence with uncompensated light mental and serious physical wounds) averages 0.3 successes. This is the primary tradeoff when considering whether to install cyberware in mages and shamans.
NOTE: The treat spell can only be used in the same mud-hour as when damage was taken, in exchange for lower drain.
NOTE: Heal/treat spells, as well as biotech treat, will work on physical drain damage.

Ritualcast now enables buffing with automatically maximized successes. The following dice optimization considerations are only applicable if you're buffing up by casting normally.

A mage's greatest strength is their ability to sustain more and higher force buff-type spells on themselves through the use of elementals.
NOTE: Elementals cannot sustain health category spells.

The effectiveness of most sustained spells is from the number of successes you achieve, where the force only serves to cap the number of successes you may use (notable exception: armor). So, it's common for spellcasters to recast such spells until they reach that maximum number of usable successes. This leads to the question: how should you split your spell pool between casting and drain resistance so as to both maximize your chance of reaching maximum effective successes while also minimizing your time spent recovering from drain?
NOTE: Drain power is calculated from half of force (round down), plus a modifier and damage level depending on the spell.

This depends on too many details to explore here. That said, when outside of combat, you can cast a spell every half second, so let's take a look at how many drain resistance dice (willpower + spell pool) it takes to have around a 90% chance to not take any drain, i.e., around 10 attempts per rest. Let's also aim for enough spellcasting dice (sorcery + spell pool + foci) for around a 10% chance of hitting the effective successes cap. These numbers assume the caster doesn't have a trauma damper. A key result is that you need much less dice if you're willing to aim for a result that is slightly less than the absolute maximum.

In Awakened Worlds CE, effective magic rating is capped at 20. But for mages, what you get for raising your real magic rating higher than 10-15 (depending on build) is relatively minor. You get an extra spell pool die for every 3 points of int+magic (where your total magic is capped at 20) and an extra rating point of foci that you can have active at the same time per 0.5 points of real magic before risking foci addiction. Note: banishing has never been worth using (except for some very niche scenarios) in any edition of Shadowrun, and is no different in the mud.

The assumption here is that you won't use a focus to sustain any spell that can instead be sustained by an elemental.

Important for Spellcasting and Combat:

• Power focus - This is by far the most efficient way to increase your spell pool (rating 4 = up to 5.5 spell pool dice, depending on existing attributes).
• Spell category focus - This adds its rating to the casting test for spells of the specified category.
• Specific spell focus - This adds its rating to the casting test for a specific spell.
• Sustaining focus (Willpower) - Assuming max effective successes, the spell force is added to your willpower, thus increasing your dice pool for drain resistance tests; also +1 spool / +3 will; also increases TN for opponent manabolt/stunbolt
• Sustaining focus (Reflexes) - This increases your initiative, allowing you to take actions (including spellcasting) faster and more often when in combat
• Spell defense focus - This adds its rating to your dice pool for spell defense

These foci could be worth having active, depending on your build:

• Sustaining focus (Body) - The extra damage resistance dice may be relatively minor vs +8 cpool from elemental sustained combat sense; beta grade dermal sheathing is more efficient at +1 bod / 0.21-0.315 magic lost vs +1 focus rating / 0.5 magic; can be useful for increasing opponent powerbolt TN, but only if you don't already have 11+ body.
• Sustaining focus (Strength) - Specifically for non-monowhip melee builds.
• Weapon Focus - For builds that rely on melee offense; not needed for melee defense (you'll already have 16 clash dice from skill+cpool, plus high armor/soak for whatever makes it through); if used for spirits, you're not casting so spell category or specific spell foci can be deactivated.

Being able to have these foci active, on top of everything above, is generally unneeded except for bragging rights, niche cases, or maybe pruns:

• Sustaining focus (Intelligence) - Cultured cerebral boosters are more efficient at +1 int / 0.2 magic lost vs +1 focus rating / 0.5 magic; +1 spool / +3 int; +1 cpool / +2 int (minor vs +8 cpool from elemental sustained combat sense); +1 hacking pool / +3 int.
• Sustaining focus (Quickness) - Cultured muscle toners are more efficient at +1 qui / 0.15 magic lost vs +1 focus rating / 0.5 magic; +1 cpool / +2 qui (minor vs +8 cpool from elemental sustained combat sense); +1 armor ratings from worn armor / +2 qui (minor vs +8 cpool from elemental sustained combat sense)

Being able to have these foci active, on top of everything above, is only really relevant for convenience or maybe pruns:

• Additional Spell Category / Specific Spell foci - You can only cast one spell a time, so you can deactivate whatever non-applicable spell category / specific spell foci as needed to avoid foci addiction.
• Sustaining focus (Charisma) - Only relevant when conjuring, so you can deactivate casting foci as needed to avoid foci addiction.
• Conjuring focus - Only relevant when conjuring, so you can deactivate casting foci as needed to avoid foci addiction.

These are not worth the space they take up, except maybe in pruns:

• Expendable foci - In the mud, there's never a case where you're limited to a single attempt (even the heal spell can be cancelled and re-cast). That makes it very difficult to justify using a non-stackable single-use consumable for a small number of dice.

Unlike a shaman, a hermetic mage's spells are limited by the quality of their library. There are no sorcery libraries with a rating greater than 8 available in Awakened Worlds CE. That restriction causes hermetic attack spells to be very unreliable when facing off against certain top tier opponents. Example: a force 8 stunbolt with 24 dice vs 16 willpower gives less than 4% chance of success, whereas a force 12 stunbolt in the same situation gives a 20% chance of success.

Where the hermetic has the advantage is their elementals, which have the ability to sustain spells associated with their element. A shaman is not able to sustain the various beneficial spells (beyond what can be handled by the available force 3 or 4 sustaining foci) without having to deal with sustaining penalties in subsequent actions. Thus, a hermetic is better at using their magic in a self-supporting role. Along with the weakness in offensive spells, it makes sense for high-tier mages to approach situations they find dangerous in the same way as a street samurai (i.e., shooting large firearms loaded with APDS from outside the room).
NOTE: A mundane street samurai can reach rating 12 with firearms and melee skills and so is generally better at killing (except when opponents are spirits/elementals). A hermetic with a full suite of buffs plus spell defense / reflecting is generally better at surviving (even more so if willing to pick up some cyber/bioware).
NOTE: In a team situation, a shaman is better at support magic, as they can sustain force 12 spells on teammates while the hermetic is limited to force 8. While the shaman would take sustaining penalties, that doesn't matter when teammates can be relied on to handle the action.

• In character generation, spell points should generally be used for spells. Once in-game, learning spells takes significantly more effort compared to buying conjuring materials or earning karma to bind foci.

• For the purpose of bonding foci, 1 spell point = 1 karma. For the purpose of learning spells, 1 spell point = 1 karma + convenience of not having to find/buy/design the spell formula (and access to a sorcery library). For the purpose of elemental materials, 1 spell point = 1k nuyen. The chargen price of 25k nuyen per point is more than a little excessive and should be avoided if possible.

• Melee is defended against by melee, so pick up a melee skill.

• Keep the illegal stuff holstered/sheathed/put away until you actually need them (NOTE: this includes the improved invisibility spell!). That way, if they're not already aggressive, you can just knock everybody out with stunbolts without entering combat.

• Because drain codes are calculated from spell force divided by 2, rounded down, it is more efficient from a drain perspective to cast spells with odd number force.

• Trauma dampers do apply to drain, which is very much worth it despite the bioware index penalizing your magic rating. Being bioware, it doesn't affect your essence and so it doesn't penalize the essence dependent spells that you might wish to cast on yourself (e.g., heal).

• Speaking of essence dependent spells, the mud does allow you to use a decrease attribute spell to reduce the TN for an increase spell for the same attribute, after which you can release the decrease spell.

• It is perfectly ok to use elementals (and thus sustained spells) at less than maximum force. Summoning force 5 elementals instead of force 8 will increase number of services by a factor of 2.4 on average. Force 5 elementals also require fewer conjuring materials, so in total, you'll be able to support 3.8x as many spells for the same amount of nuyen. Just force 5 combat sense (with 10 spellcasting successes) and force 5 armor are enough to put your combat pool and armor ratings on-par with or better than most street samurai.

• The odds of a mundane elite guard with 8 dice beating the spell resistance check vs a force 5 imp invis (with 5 spellcasting successes) is 9%. Against force 6 (with 6 successes), their odds are 0.04%. This means if you know you're walking into a fight, go ahead and cast that (and sustain with a water elemental) because it means everybody (except for dual-natured creatures) suffers TN penalties against you (+2 to shoot/melee/dodge if astrally perceiving, at least +4 if ultrasound, +8 for everybody else).

• Autorun Johnsons ignore stealth spells (as well as invis spells). The stealth spell also does not cause guards to aggro, so it's perfectly ok to keep it active at all times. If you use the stealth skill, the spell makes it much less likely for opponents to beat the perception check. NOTE: sneaking works when moving between rooms; stand in a room long enough and opponents will spot you without needing a perception check.

• When fighting against certain high force spirit that also cast spells, allocate your spool to defense/reflecting and use a weapon focus.

• Against particularly strong critters with particularly high attributes, bring a large firearm loaded with APDS and shoot in from outside the room.