From the FreeBSD 13.0 manual page:

DESCRIPTION
     The sysctl utility retrieves kernel state and allows processes with
     appropriate privilege to set kernel state.  The state to be retrieved or
     set is described using a “Management Information Base” (“MIB”) style
     name, described as a dotted set of components.

This description should probably emphasize the valuable information that can be read from sysctls:

hw.ncpu: Number of active CPUs - Often used with for the make(1) -j flag
hw.physmem: Amount of physical memory (in bytes) - The physical RAM of the system
kstat.zfs.misc.arcstats.size: size - The current size of the ZFS ARC read cache
kstat.zfs.misc.arcstats.c_max: c_max - The current maximum size of the ZFS ARC read cache
kstat.zfs.misc.arcstats.c_min: c_min - The current minimum size of the ZFS ARC read cache
vfs.zfs.arc_max: max arc size (LEGACY) - The writable maximum size (FreeBSD 12 and before)
vfs.zfs.arc_min: min arc size (LEGACY) - The writable minimum size (FreeBSD 12 and before)
vfs.zfs.arc.max: Max arc size - The writable maximum size (FreeBSD 13 and later)
vfs.zfs.arc.min: Min arc size - The writable maximum size (FreeBSD 13 and later)

A list of all available sysctls with descriptions can be produced with:

sysctl -ad

kern.ostype: Operating system type
kern.osrelease: Operating system release
kern.osrevision: Operating system revision
...

The 13,000+ descriptions range in verbosity and consistency.

As a facility for built-in, developer-facing documentation, the quality of the FreeBSD sysctl descriptions has a direct impact on the FreeBSD developer experience. When authoring a new sysctl description, please consider these guidelines:

Please do not leave the sysctl -d output blank. Some categories have descriptions while many do not.

vm.domain: 
vm.domain.0: 
vm.domain.0.pidctrl: 
vm.domain.0.pidctrl.kdd: Inverse of derivative gain

kstat.zfs: 
kstat.zfs.zroot: 
kstat.zfs.zroot.dataset:

A full list of missing descriptions can be obtained with:

sysctl -ad | grep '[: ]$'

External components like OpenZFS and drivers can be a challenge.

hw: hardware
hw.machine: Machine class
hw.model: Machine model

net: Network, (see socket.h)
net.local: Local domain
net.local.stream: SOCK_STREAM

kern.geom.disk: GEOM_DISK stuff
kern.geom.disk.ada1: GEOM disk ada1
kern.geom.disk.ada1.flags: Report disk flags

vfs.zfs: ZFS file system
vfs.zfs.sync_pass_rewrite: Rewrite new bps starting in this pass
vfs.zfs.sync_pass_dont_compress: Don't compress starting in this pass

Suggestion: Survey these and propose a format to follow

Suggestion: Do not use “stuff”

Suggestion: Consider manual page references

hw.dri.debug: Enable debug output, where each bit enables a debug category.
                Bit 0 (0x01)  will enable CORE messages (drm core code)
                Bit 1 (0x02)  will enable DRIVER messages (drm controller code)
                Bit 2 (0x04)  will enable KMS messages (modesetting code)
                Bit 3 (0x08)  will enable PRIME messages (prime code)
                Bit 4 (0x10)  will enable ATOMIC messages (atomic code)
                Bit 5 (0x20)  will enable VBL messages (vblank code)
                Bit 7 (0x80)  will enable LEASE messages (leasing code)
                Bit 8 (0x100) will enable DP messages (displayport code)

hw.ixl.enable_vf_loopback: 
Determines mode that embedded device switch will use when SR-IOV is initialized:
        0 - Disable (VEPA)
        1 - Enable (VEB)
Enabling this will allow VFs in separate VMs to communicate over the hardware bridge.

hw.ixl.i2c_access_method: 
I2C access method that driver will use:
        0 - best available method
        1 - bit bang via I2CPARAMS register
        2 - register read/write via I2CCMD register
        3 - Use Admin Queue command (best)

machdep.syscall_ret_flush_l1d: Flush L1D on syscall return with error (0 - off, 1 - on, 2 - use hw only, 3 - use sw only

vfs.zfs.vol.mode: Expose as GEOM providers (1), device files (2) or neither

vm.background_launder_max: background laundering cap, in kilobytes
vm.background_launder_rate: background laundering rate, in kilobytes per second

vm.stats.vm.v_page_size: Page size in bytes

External components can be a challenge. Determine if the output should be fixed upstream or patched.

sysctls with limited options benefit from listing those options.

Providing the input unit is helpful such as “in kilobytes per second” or “in bytes” is very helpful.

kern.random.random_sources: List of active fast entropy sources.
kern.random.initial_seeding.disable_bypass_warnings: If non-zero, do not log a warning if the 'bypass_before_seeding' knob is enabled and a request is submitted prior to initial seeding.
kern.random.initial_seeding.arc4random_bypassed_before_seeding: If non-zero, the random device was bypassed when initially seeding the kernel arc4random(9), because the 'bypass_before_seeding' knob was enabled and a request was submitted prior to initial seeding.

vm.pmap.large_map_pml4_entries: Maximum number of PML4 entries for use by large map (tunable).  Each entry corresponds to 512GB of address space.

Suggestion: Rephrase description sentences into “headlines” with units and references as appropriate.

150+ descriptions of 13,000+ end in periods. A full list of descriptions ending in periods can be obtained with:

sysctl -ad | grep '[.]$'

Consider avoiding characters such as single quotation marks and unnecessary parentheses that may be a challenge to machine parsing:

kern.ps_strings: Location of process' ps_strings structure

kern.sched.steal_thresh: Minimum load on remote CPU before we'll steal

kern.vt.kbd_reboot: Enable reboot keyboard combination.  See kbdmap(5) to configure (typically Ctrl-Alt-Delete)

Consider a format that includes a brief description, the units, if writable, and a reference:

category.foo.thing: Writable kernel foo count in bytes, writable. See foocount(5)

or

category.foo.thing: Writable kernel foo count - bytes - writable - foocount(5)

Suggestion: Consider a having a goal of generating tabular output with this criteria with a suitable separator.

Suggestion: new kernel build option SYSCTL_IMPL_DESC, if it is enabled add _ _FILE_ _ to the description to know the entry point/implementation file of the object. This proposal needs a cost benefit analysis; possible implementation: sysctl(9) macros or raw SYSCTL_OID().