What I dreamed up isn't really a roguelike (oops) or a tower defence game
(oops again). It's more like "Dwarf Fortress done right"; holding out
against invaders while managing resources rather than a mess of
micromanagement and absurd simulationism. In particular, like most TD
games, there's no really defined technological basis. Guards have ranged
weapons, melee weapons, and armour. What are they? Who cares?

It's also quite shamelessly "inspired" by Stronghold, the videogame.

In view of this, your fortress consists of towers and walls, but most of
your defensive firepower is in guards, independently mobile units with a
simple AI. Towers contain facilities such as armourers, barracks, hospitals
or storage rooms - or mounted heavy weapons - and guards can fire out from
towers with relative impunity. However, if towers aren't close together
(and they are expensive to build), guards will need to patrol the walls
and repel the creeps who will otherwise swarm across them or destroy them
when out of range of towers.

Your initial tower is special only inasmuch as one of the facilities it
contains is a "portal". A portal can supply you with guards (at a price),
food (at a price), other resources (such as stone, wood, iron,
manufactured weapons and armour); it can also be used to sell surplus
resources. Ideally automated instructions would avoid micromanagement
here. Money is also earned by killing creeps; whoever's on the other side
of the portal will pay you money for holding off the barbarians. The basic
creep AI seeks to get to your portal (over, through, or around
fortifications, as their temperament takes them) and destroy it; if your
portal (portals? They might be constructable at ruinous expense) is
destroyed, they will seek to destroy other towers, and in practice if you
can't build a fresh one, the game is now lost.

The fundamental resource gathering mechanic is to enclose territory. By
default, territory enclosed by your walls is farmed, producing food that
can support guards or be sold. You might also enclose stone or iron
outcroppings, or forests, or other mineable resources, which could then be
extracted. If creeps get over the walls, squares they touch do not produce
food for a while?

Initially I figured you'd just place structures in hostile territory like
any TD game, but you might get clever with a need to make a sally or scout
approaching waves in order to carry out construction without coming under
fire. Regardless I would suggest that a fortress without an exterior gate
can't be expanded, and that interior gates at ground level are necessary
for areas to become farmed, or to do certain kinds of activity in towers
(you can't practically run an ironworks up an itty-bitty staircase). This
will provide weak points for attack.

Territorial effects could also keep the map from being the usual blank
canvas. Beyond the minable resources, territory like a swamp (retards
movement, ruinously expensive to build in), impassible river (actually
helpful), chasm (unbuildable, spawns creeps), too-high cliff (effectively
void space that makes the game easier) would liven things up.

Facilities could be built either inside towers (which are expensive) or up
against walls (which is less efficient territory-wise because towers are
multistorey). Ones that came to mind are portals, barracks, medical
facilities, training facilities, recreation (especially if you have any
tracking of morale), storage facilities, weapon and armoursmiths, resource
mines, ironworks, some kind of research facility to unlock better weapons,
armour, stronger walls... but remember the idea is not to be Dwarf
Fortress. Detail should have a meaningful game mechanical effect.

Your guards are probably organised into three shifts. One shift is on
duty; one relaxing or training; one asleep. Each tower has an assigned
number of guards; they patrol the walls around the tower, man the tower's
weapons, or fire from arrowslits in the tower, as configured by the
player. Remaining on-duty guards (if any) form the "reserve". When a
tower's guards engage in combat, a configurable fraction of the reserve is
dispatched to that tower, and the player can also manually dispatch chunks
of the reserve to any tower under attack. In serious situations, the
player can order the off-duty shift to join the reserve, or even the
sleeping shift - but this will have long-term consequences in terms of
fatigue and disruption of training.

I wouldn't get into complicated injury modelling, but you might have a few
injury states for guards; unhurt, scratched (does not affect fighting
ability), injured but fighting, walking wounded (can still just about
defend themselves), downed, and dead. The player should also be able to
control at what state a guard will be replaced with someone from the
reserve - and at what state they will bug out immediately rather than
waiting for their replacement - and at what state the guard should remove
themselves from the duty roster altogether. Then in an emergency you might
also call up injured-but-fighting guards and send them to the walls;
conversely, a well-run fortress might take anyone who gets a scratch off
duty until they heal.

The real trick here should be to make the decisions meaningful. Should I
enclose a lot of land with a flimsy defence, accept that creeps will
damage it and that my guards will be spread thin, but grow a lot of
valuable food and use the income for replacements - or should I enclose a
minimal amount of land with a sturdy defence and not need much income for
replacements? Is it better to have lots of cheap guards or an elite few?
Scouting facilities could provide a limited wave browsing facility (and if
the waves aren't all-over-the-place random but drift through the
possibilities, so that a wave of zombies is more likely to be followed by
another mass stupid undead attack than three tridudes with plasma rifles,
this information might be more useful); facing mass attacks, a bunch of
warm bodies who can discharge bullets in "to whom it may concern" mode is
probably more useful than three expert marksmen. Is it worth building lots
of hospital facilities, or should I save the money and accept that some
wounded men will die? If troops' initial quality depends on the wages
offered but they gain experience, a cheap "more where they came from"
strategy could work well for a while but ultimately demand you spent the
money on something useful to stand in for the lack of veterans.

I would suggest some kind of morale system. Morale increases from things
like surviving fights unhurt, being able to leave one's post when
injured, being well paid and well fed, associating with guards with high
morale or with veterans, having high quality weapons and armour - it
decreases from the opposite effects. Morale has two effects; with low
morale, troops may desert their post without orders when hurt and flee,
and if you can't meet your wage bill or food requirements, a guard with
low morale may leave your service. With high morale, their morale drops,
but they will tough it out and see if things improve. Obviously a shortage
of food has more direct effects on guard performance as well!

Essentially, I envisage the cheap warm-body strategy being attractive and
easy (and easy to fall back on in a crisis - the very lowest quality
guards might be entirely free if you have weapons and armour taken from
the fallen - perhaps they are in legal trouble on the other side a la
French Foreign Legion), leading to rapid expansion - but also fragile,
prone to spectacular collapses. A high quality strategy will in theory be
more robust but will lead to constant resource worries as you try to buy
high quality stuff on a tight budget.