Catalog gap analysis — 2026-09-30

A deep dive over every category in catalog/categories.yaml: what the catalog carries, what the wider field has that it does not, what the project planned and has not yet delivered, and where a configuration or a page would do more than another manifest. Prompted by the maintainer's own find, 2026-09-28: the GPS and navigation software was missing, and it became D-057 and D-061. This page asks what else is missing the same way, one category at a time.

What this page is. A measured list of candidates and gaps with a recommendation on each, written for the maintainer to rule on. Nothing on it is decided; the decisions it asks for are collected as Q-022 in docs/QUESTIONS.md.

What this page is not. Not a disposition (docs/reference/dispositions.md is the index of the six sources' units, and this page does not change it); not a claim that any candidate installs on any target — see How it was measured, which says exactly what was and was not checked; and not a re-litigation of anything in docs/DECISIONS.md. The RF-security group (Wi-Fi, Bluetooth, cellular, RFID, sub-GHz) is deliberately out of this pass except for two plan-versus-delivery notes: SCOPE.md's post-1.0 tracks and D-034/D-046 already govern it, and a comparison of that tooling is a separate job.

Headline

The catalog is broad where the six sources are broad. The Debian Hamradio Blend is carried whole — every backticked package in docs/reference/blend-inventory.md resolves to a manifest or to an apt block. Where the catalog is thin is not where the sources are thin; it is where the engine's own plans outran the manifests:

Gap	Planned by	Delivered
Station config consumed by the units that need a callsign	D-004, D-008 ("Direwolf with configuration"), D-035	One manifest (linbpq) carries a config_files block. Direwolf, pat, WSJT-X, fldigi, JS8Call, Xastir, YAAC, gpredict and the loggers all still ask the operator to type the callsign a second time
A rig as station data, one CAT connection shared by every program	D-042 (the ETC rig model, "reimplemented as catalog data")	Not started. Station config has no rig field; nothing runs rigctld
GPS-disciplined time on the field laptop	The station profile's own summary ("time, position")	gpsd is in station; nothing feeds it to the clock. FT8 without the internet — the D-057 scenario — fails at ±1 s
Audio routing for digital modes	The documentation standard in CLAUDE.md ("fresh Parrot to a working digital-modes station without asking anyone")	No unit, no guide, no troubleshooting entry. Every digital mode needs it
Kismet	rf-security profile prose: "held back until the third-party apt repo path is implemented"	D-040 implemented that path on 2026-09-03. Kismet is still absent and the prose is stale
The task-oriented guides	CLAUDE.md's docs/guides/ list: digital modes, APRS, satellite, packet, SDR	Four guides exist: conference operating, offline navigation, propagation, Rayhunter. None of the five named ones

Those six are the findings that change what the project does next. The per-category candidates that follow are real, but most are one manifest each; these are the ones an operator hits on the first evening.

How it was measured

• The catalog: 266 manifests read with the loader's own YAML, tagged against the 56-tag vocabulary; status and config_files counted; system_modifications kinds counted (two apt_pin, two group_membership — nothing writes a service or a config file except linbpq).
• The record: SCOPE.md, DECISIONS.md D-046 through D-061, not-carried.md, parity-coverage.md's outstanding table, prior-art.md §4 and §5, and the seven open GitHub issues.
• Upstreams: web search on 2026-09-30 for each candidate's current release, licence and activity. Where the result names a date it is the search result's date, not a head-commit check; D-032's rule (the default branch's head commit, never updated_at) has not been applied to any candidate here, and must be before any manifest is written.
• Archives: this session could not reach packages.debian.org, sources.debian.org, tracker.debian.org or Launchpad (egress policy), and the seven-target apt-cache policy sweep needs Podman. So every "in Debian" claim below is from a web search result and is marked unmeasured until scripts/apt-policy-sweep.sh --all says otherwise. A candidate is not a manifest until that sweep has run for it (D-014, D-025).

Measured since, 2026-09-30 — the apt adds of D.6

scripts/apt-policy-sweep.sh, one target at a time, over the eight apt candidates of item 6 in section D. - means the archive has no candidate.

Package	Debian 13	Ubuntu 26.04	Ubuntu 24.04	Kali	Parrot 7	Mint 22.3	Debian 13 arm64
predict	-	-	-	-	-	-	-
gr-fosphor	3.9~git20240323.74d54fc-1+b10	3.9~git20240323.74d54fc-1ubuntu1	3.9~git20230826.e02a2ea-1build3	3.9~git20240323.74d54fc-2+b1	3.9~git20240323.74d54fc-1+b10	3.9~git20230826.e02a2ea-1build3	3.9~git20240323.74d54fc-1+b10
libiio-utils	0.26-2	0.26-2build2	0.25-4build2	0.26-2+b2	0.26-2	0.25-4build2	0.26-2
stm32flash	0.7-1	0.7-1build1	0.7-1	0.7-1	0.7-1	0.7-1	0.7-1+b1
ser2net	4.6.4-1	4.6.5-1	4.6.0-1build2	4.6.5-1	4.6.4-1	4.6.0-1build2	4.6.4-1
qpwgraph	0.8.2-1	0.9.9-1	0.6.1-1build3	1.0.3-1	0.8.2-1	0.6.1-1build3	0.8.2-1
gpsprune	25.2-1	26.1-1	23.2-1	27+ds-1	25.2-1	23.2-1	25.2-1 (all)
opencpn	1:5.10.2+dfsg-1	1:5.12.4+dfsg-1	5.8.4+dfsg-1build4	1:5.14.1+dfsg-4	1:5.10.2+dfsg-1	5.8.4+dfsg-1build4	1:5.10.2+dfsg-1

What that changed below, each noted at its row:

• predict is in no target's archive, and not in Debian 12 or unstable either (apt-cache search in debian:12, debian:13 and debian:sid, 2026-09-30). The search result's "in testing" was wrong. Debian's packaging repository (salsa debian-hamradio-team/predict) was last touched on 2018-12-30, with an upload marked UNRELEASED. Not added; satellite-tracking stays a one-member category until a maintained route is found.
• gr-fosphor is offered on all seven targets and is added, but it is not a Blend sdr task member: blend-inventory.md, generated from the task files, does not list it. The row below said it was.
• stm32flash: no owned device is recorded as using the STM32 serial bootloader. The hardware catalog's STM32 entries (flipper-zero, the dmr-radio class) use USB DFU (0483:df11), which is dfu-util's job; NanoVNAs, named in the row below, are not catalogued as devices at all. The unit is added and links to no device.
• ser2net starts a service on install. Read from the Debian 13 .deb and measured in a container: the postinst enables ser2net.service, which runs as root, and the shipped configuration listens on loopback ports 2000, 2001, 3000 and 3001 for /dev/ttyS0 and /dev/ttyS1. Added, but not to station as the brief placed it: install it by name.
• gpsprune is offered on all seven targets and is not added, because the decision record already ruled on it: D-061 (and the 2026-09-28 note under D-057) says GPSPrune was measured and is not carried, because it uses online tiles, and the navigation profile's exclusions say the same. This report recommended it without citing that. Carrying it would need the maintainer to amend D-061; nothing here does.

Measured since, 2026-10-01 — the first source-built adds of D.10

scripts/apt-policy-sweep.sh, one target at a time, over the three units' names and their build dependencies. - means the archive has no candidate.

Package	Debian 13	Ubuntu 26.04	Ubuntu 24.04	Kali	Parrot 7	Mint 22.3	Debian 13 arm64
freedv	1.8.11-1+b5	1.8.11-1build5	1.8.11-1build4	1.8.11-1+b5	1.8.11-1+b5	1.8.11-1build4	1.8.11-1+b5
libcodec2-dev	1.2.0-3	1.2.0-4	1.2.0-2build1	1.2.0-5	1.2.0-3	1.2.0-2build1	1.2.0-3
librade0, librade-dev, radae, rade	-	-	-	-	-	-	-
pihpsdr, deskhpsdr, libwdsp-dev	-	-	-	-	-	-	-
droidstar, dudestar, mvoice, md380-emu	-	-	-	-	-	-	-
libgtk-3-dev	3.24.49-3	3.24.52-0ubuntu1	3.24.41-4ubuntu1.3	3.24.52-1	3.24.49-3	3.24.41-4ubuntu1.3	3.24.49-3
libgpiod-dev	2.2.1-2+deb13u1	2.2.1-3build1	1.6.3-1.1build1	2.2.5-1+b1	2.2.1-2+deb13u1	1.6.3-1.1build1	2.2.1-2+deb13u1
qt6-base-dev	6.8.2+dfsg-9+deb13u2	6.10.2+dfsg-7	6.4.2+dfsg-21.1build5	6.10.2+dfsg-16	6.8.2+dfsg-9+deb13u2	6.4.2+dfsg-21.1build5	6.8.2+dfsg-9+deb13u2
libwxgtk3.2-dev	3.2.8+dfsg-2	3.2.9+dfsg-1	3.2.4+dfsg-4build1	3.2.11+dfsg1-1	3.2.8+dfsg-2	3.2.4+dfsg-4build1	3.2.8+dfsg-2

Every other build dependency of the three (the rest of piHPSDR's list, Qt 6's multimedia, declarative and serial-port packages and the QML modules, FreeDV's audio and wxWidgets stack) has a candidate on all seven. What that changed below:

• piHPSDR is carried (catalog/packages/pihpsdr.yaml, sdr), from DL1YCF's repository, the maintained one, at tag v3.0. Its licence is GPL-3.0-or-later, not the GPL-3.0 the row below gives without the "or later". Built in Debian 13 and Ubuntu 24.04 containers.
• DroidStar is carried (catalog/packages/droidstar.yaml, digital-modes) on Debian 13, Parrot, Kali and Ubuntu 26.04. Ubuntu 24.04 and Mint carry Qt 6.4, below its 6.5 floor. Its downloadable vocoder plugin, the reason this report flagged it, was removed upstream on 2025-10-19; its build turns on an ARM-only MD-380 firmware vocoder, which a two-line patch turns off. Not the .deb/AppImage route the row below suggests: those are third-party builds, not measured.
• FreeDV 2.x is not carried yet. Every target has 1.8.11. v2.4.0 builds, but its CMake fetches RADE's C port and RNNoise from main while building, which cannot be pinned from outside (source-build-gaps.md #9, not-carried.md).

Measured since, 2026-10-01 — the second source-built batch of D.10

scripts/apt-policy-sweep.sh, one target at a time, over the six units' names, the names their neighbours go by, and their build and runtime dependencies. - means the archive has no candidate.

Package	Debian 13	Ubuntu 26.04	Ubuntu 24.04	Kali	Parrot 7	Mint 22.3	Debian 13 arm64
dump978, dump978-fa, skyaware978, uat2tools	-	-	-	-	-	-	-
nrsc5, libnrsc5-dev	-	-	-	-	-	-	-
librevna, librevna-gui	-	-	-	-	-	-	-
k5prog, radio-tool, radiotool	-	-	-	-	-	-	-
tar1090, dump1090-fa	-	-	-	-	-	-	-
readsb	3.14.1630+git20240609.adc080d-1	3.16-2build1	-	3.16-2	3.14.1630+git20240609.adc080d-1	-	3.14.1630+git20240609.adc080d-1
lighttpd	1.4.79-2	1.4.82-2ubuntu2	1.4.74-1ubuntu3	1.4.85-1	1.4.79-2	1.4.74-1ubuntu3	1.4.79-2
libfaad-dev	2.11.2-1	2.11.2-1build1	2.11.1-1build1	2.11.3-1	2.11.2-1	2.11.1-1build1	2.11.2-1
libboost-program-options-dev	1.83.0.2+b2	1.90.0.1ubuntu3	1.83.0.1ubuntu2	1.90.0.2+nmu1	1.83.0.2+b2	1.83.0.1ubuntu2	1.83.0.2+b2
libsoapysdr-dev	0.8.1-5+b2	0.8.1-7build1	0.8.1-4build1	0.8.1-7+b2	0.8.1-5+b2	0.8.1-4build1	0.8.1-5+b2
qt6-base-dev	6.8.2+dfsg-9+deb13u2	6.10.2+dfsg-7	6.4.2+dfsg-21.1build5	6.10.2+dfsg-16	6.8.2+dfsg-9+deb13u2	6.4.2+dfsg-21.1build5	6.8.2+dfsg-9+deb13u2
qt6-svg-dev	6.8.2-3	6.10.2-2	6.4.2-4ubuntu3	6.10.2-9	6.8.2-3	6.4.2-4ubuntu3	6.8.2-3

Every other build dependency of the six (the toolchain, cmake, autotools, libao-dev, libfftw3-dev, librtlsdr-dev, libusb-1.0-0-dev, qmake6, libgl-dev) has a candidate on all seven. libfaad-dev is offered everywhere but is the unpatched FAAD2; nrsc5 builds its own. What that changed below:

• dump978-fa is carried (catalog/packages/dump978-fa.yaml, listening), FlightAware's tag v11.1, built on Debian 13 and Kali. Its licence is BSD-2-Clause. It decoded upstream's own sample to 21 aircraft, and Debian's readsb took the same sample as a second input.
• nrsc5 is carried (catalog/packages/nrsc5.yaml, listening), tag v3.2.0. Its CMake clones FAAD2 at a tag while building; a two-line patch makes that a sha256-checked archive (source-build-gaps.md #10).
• LibreVNA is carried (catalog/packages/librevna.yaml, electronics), tag v1.6.5, built on Debian 13, Ubuntu 24.04 (Qt 6.4) and Kali. Its SCPI server listens on every address by default, which its page says. The device entry (catalog/hardware/devices/librevna.yaml) is not owned and its identifiers come from upstream's source.
• k5prog is carried (catalog/packages/k5prog.yaml), in no profile, at its untagged head from 2023. Its licence is GPL-3.0-or-later, not the GPL the row below gives.
• radio_tool is carried as radio-tool (catalog/packages/radio-tool.yaml), in no profile. It is v0l's repository, which OpenRTX's guide points at; there is no OpenRTX copy. The pin is past the last tag, which cannot flash, and Bluetooth LE is switched off so the build fetches nothing.
• tar1090 is added (2026-10-02) (catalog/packages/tar1090.yaml, listening; depends on readsb). Its page works from any static server on loopback over readsb's JSON, so the route was a reference serve page, not lighttpd: /aircraft/, reading /run/readsb, with the pinned archive's html/ (commit e784ee5, own-choice pin, D-024) installed as data. It cannot call out: its fetching settings are off, its online map layers are replaced by the station's PMTiles map (a blank background and the reason on the page without one) and the page's responses carry a policy naming no host. Its aircraft database cannot be pinned and is not carried (dispositions.md).

A. Cross-cutting gaps — plans the catalog has not caught up with

A1. Station config is built and almost nobody reads it

hammunition station set stores callsign, grid square and node alias at mode 0600 (D-035), and the plan defers a file whose value is missing rather than refusing the unit. Exactly one manifest uses it. These are the units whose first run asks for the same two values, with where each keeps them. Every one is a candidate config_files block; the third column is the risk that decides whether a block is right or a page is:

Unit	File and keys	Shape	Risk
direwolf	direwolf.conf: MYCALL, ADEVICE, PTT	Whole file from the shipped example, callsign filled	Low. D-008 promised this by name. Audio device and PTT stay the operator's; a file with MYCALL set and the rest commented is still the right starting point
pat	config.json: mycall, locator, secure_login_password	Merge into JSON	Medium: the password must never be templated (it is not station data and is not stored), and pat rewrites its own file
ax25-tools	/etc/ax25/axports: the wl2k ${MYCALL} line 73Linux writes	Append (append: true exists in the schema)	Low, and already named in CLAUDE.md as the case that forced D-004
xastir, yaac	Xastir's config/xastir.cnf (STATION_CALLSIGN, STATION_LAT/LONG); YAAC's YAAC.conf (callsign, latitude, longitude)	Whole file on first run only	Medium: both rewrite on exit; backup_existing covers the first run, a second install must not clobber
wsjtx, wsjtx-improved, jtdx, mshv, js8call	WSJT-X.ini etc.: MyCall, MyGrid under [Configuration]	Merge into INI	High. Qt writes the whole file on exit and the first-run wizard exists for a reason. Recommend a page, not a block: the settings dialog with the two fields named
fldigi (and the fl* family)	fldigi_def.xml: <MYCALL>, <MYQTH>, <MYLOC>, <MYNAME>	Merge into XML	High, same reason
gpredict	~/.config/Gpredict/ QTH file: LAT, LON from the grid square	Whole file	Low. The grid-to-lat/lon conversion is four lines and the engine already has a locator
cqrlog, klog, qlog, xlog, pyqso, tlf, not1mm	Each its own; tlf's logcfg.dat is plain text (CALL=, MYLOCATOR=)	tlf whole file; the GUI loggers a page	Low for tlf; the others run a wizard
aprx, uronode, linpac, fbb	Plain-text configs with a callsign line	Whole file	Low; same class as linbpq
hamclock-next, openhamclock	Grid square on first run	A page	The value is entered once in a GUI

Recommendation. Write the blocks for the low-risk plain-text units (direwolf, ax25-tools, gpredict, tlf, aprx, uronode, linpac, fbb), in that order, each behind the D-035 deferral. For the Qt and XML units, a page: the digital-modes guide (A9), section "Enter your callsign and grid once", with the dialog path for each program. Do not template a password anywhere; pat's block sets mycall and locator and leaves the password key absent, and the plan says so.

Measured, 2026-09-29 (Q-022 #1, ruled yes). The table above was written from memory and search results; building the blocks measured each unit's shipped files, man page or source, and three rows were wrong. tlf's locator key is MYQRA: the packaged binary refuses MYLOCATOR. linpac gets no block: its first-run questions run only while ~/LinPac is absent, and four of the five are not station values. fbb gets no block: fbb.conf's mandatory lines include the hierarchical address, SSID, city and sysop name, so a block could only write the partial file D-035 forbids. The other six carry blocks -- direwolf (/etc/direwolf.conf), ax25-tools (/etc/ax25/axports, appended once), gpredict (~/.config/Gpredict/sample.qth), tlf (~/tlf/logcfg.dat), aprx (/etc/aprx.conf) and uronode (/etc/ax25/uronode.conf) -- and each manifest cites what was measured and says what was not. The engine work they needed is D-035's 2026-09-29 amendment.

A2. The rig is not station data, so every program is configured separately

D-042 approved reimplementing ETC's rig model as catalog data and named it sub-project work. Nothing has started. What the model needs, in this project's shape:

1. A rig hardware class (catalog/hardware/classes/), alongside gps-receiver and dmr-radio: the USB-serial CAT interface, its dialout membership, and the D-028 warning that a CP2105 or FTDI identifier names a chip, not a radio. The maintainer's FT-991A enumerates as a CP210x pair (audio on a separate USB audio class device).
2. Station fields: rig (a hamlib model number or name), rig_device (a /dev/serial/by-id/ path, never /dev/ttyUSB0), rig_baud. All three are D-035 values: missing defers, nothing guesses.
3. One rigctld, shared. libhamlib-utils gains a systemd user service (~/.config/systemd/user/, scope: user) running rigctld -m <rig> -r <device> -s <baud>, and every CAT-capable unit's page says "Hamlib NET rigctl, localhost:4532". flrig is the alternative for operators who prefer it and is already carried; the guide states the trade-off (flrig has the panel, rigctld has no UI and no conflicts).
4. ser2net (apt, unmeasured) for the rig on another machine, the same service in reverse. One unit, serial-terminals.

Recommendation. Do 1 to 3 as one sub-project after A1, since the service is a config_files block plus a system_modifications entry the schema does not yet have (a user service is neither a group nor a pin). Record the new kind in D-035's amendment.

A3. Audio routing has no unit, no guide and no troubleshooting entry

Every digital mode needs the radio's audio in and out of the program, and on every 2026 target that is PipeWire. What is missing:

• A patchbay. qpwgraph (apt on every target that ships PipeWire; unmeasured) is the program an operator uses to see which device WSJT-X is actually listening to. pavucontrol is the fallback. One unit, tagged device-support, in digital-modes and station.
• Virtual cables. SDR-to-decoder (SDR++ audio into WSJT-X, gqrx into fldigi) needs a loopback: pw-loopback or a libpipewire-module-loopback stanza under ~/.config/pipewire/pipewire.conf.d/. That is a config_files block with scope: user — and a known failure to record: JS8Call-improved and Qt6 WSJT-X builds do not list PipeWire virtual devices as inputs on some systems (upstream issue #120, reported on Arch with PipeWire), with the ALSA snd-aloop route as the workaround.
• The audio level page. Direwolf's manifest already says transmit level is the biggest cause of a station that is heard by nobody; the same is true of every mode and nothing says it in one place.

Recommendation. A new guide, audio-routing, under docs/guides/, first (it is the page the documentation standard requires), then the qpwgraph unit, then the loopback block as a digital-modes profile option once the Qt6 input bug has been measured on the field laptop.

A4. GPS time: gpsd is installed and the clock ignores it

The station profile's summary is "rig control, time, position", and it installs gpsd. Nothing connects gpsd to the system clock. On the field laptop with the network down — the case D-057 was written for — FT8, FT4, JS8 and WSPR all need the clock within about a second, and a laptop's RTC drifts past that in days.

• chrony with refclock SHM 0 refid GPS (or the SOCK refclock, which chrony's own documentation prefers over SHM for permissions) and refclock PPS /dev/pps0 lock GPS where the receiver has PPS. Debian 13 installs systemd-timesyncd by default; whether Parrot, Kali and the Ubuntu targets ship chrony already is unmeasured and matters (D-022: coexist and disclose, never remove silently — replacing timesyncd is a disclosed system modification).
• The receiver itself: ubxtool is already carried in gpsd-clients. PyGPSClient (PyPI, BSD-3, tkinter; the u-center replacement that configures u-blox receivers' output protocols and rates, plus NTRIP for RTK) is a venv unit for the gps-gnss tag.
• The field laptop's planned GNSS module (CLAUDE.md: not fitted) is the measurement rung; a USB puck is the meantime. docs/reference/bench-verification-5430.md is where the result goes.

Recommendation. A chrony manifest with one config_files block and a system_modifications entry naming the timesyncd displacement, in station; a troubleshooting entry "my FT8 decodes nothing: check the clock" in docs/troubleshooting/running.md.

A5. The terrain layer D-061 built serves one reader and could serve four

D-061 fetches Copernicus GLO-30 tiles, hashed and verified, for QMapShack's contours. Three carried units want the same elevation data and get none:

Unit	What it reads	Route from the GLO-30 tiles
splat	SDF files made by srtm2sdf / srtm2sdf-hd from SRTM .hgt	gdal_translate -of SRTMHGT per tile, then srtm2sdf-hd; a derived converter (splat-sdf) beside gdal-dem
xastir	GeoTIFF or DRG topo layers	GDAL already carried (gdal-bin); a derived converter producing the tiled GeoTIFF Xastir reads
signal-server (candidate, below)	SPLAT's SDF	Same converter as splat
gpredict, hamclock-next	Nothing — but a horizon mask from the DEM is what a satellite station actually wants	Post-1.0

Recommendation. One converter, splat-sdf, in the derived method's enum, and splat and signal-server become readers of the station's regions the way QMapShack is. It closes "propagation prediction with no network", which the propagation profile page currently cannot promise.

Built 2026-10-02 (D-061's amendment of that date): splat-sdf and signal-server, in antenna, where splat is. Measured on the way: Cloud-RF's repository now holds only a README, so signal-server is W3AXL's fork; Xastir reads a GeoTIFF or a .geo-described image, not an SDF, so the converter does not serve it, and its route is recorded in the amendment; SPLAT's -geo output is an Xastir layer meanwhile.

A6. Offline reference data the D-049 shape can now carry

D-049 gave the catalog a data method; D-057 gave it regional and dated fetches. country-files (cty.dat) is carried. Two more belong in the same shape and are missing:

• The FCC amateur licence database. l_amat.zip from data.fcc.gov/download/pub/uls/complete/, regenerated every Sunday, public domain, on the order of 200 MB. Every carried logger does its lookups against QRZ or HamQTH over the network; with this file on disk, a US station can look up any callsign with the network down. The file's freshness is D-057's weekly case, not a pin. What reads it is the open question: cqrlog and klog have no offline-callbook importer, so the first reader would be a small tool of ours (references) or a documented sqlite recipe.
• Band plans and a repeater directory. An ARRL band-plan chart is data with a licence to check. RepeaterBook's export is not freely licensed and cannot be a data unit; the retired 73Linux REPEAT entry was a client for it, and that ruling stands. Note the gap, carry nothing.

A7. Kismet: the blocker moved and the profile did not notice

catalog/profiles/rf-security.yaml says Kismet "is held back until the third-party apt repo path is implemented rather than shipped as a profile that half-installs." D-040 implemented that path on 2026-09-03, with code and codium as its exercised cases. Kismet is apt on Kali and Parrot, and elsewhere comes from its own signed repository with Debian 13 packages (the project refreshed its signing key in February 2026 for the new apt signature requirements — the fingerprint in a manifest must be the current one, read from the key, per src/hammunition/openpgp.py). The maintainer's CatSniffer V3 is a Kismet datasource (Zigbee and BLE), so the device page and the unit belong together.

Recommendation. A kismet manifest: apt block selected for Kali and Parrot, apt_repos block elsewhere, rf-security profile; update the profile's deliberately_excludes prose in the same commit.

A8. Waiting items whose blocker has changed since they were written

Item	Recorded as	What changed	Recommendation
ARDOPGUI	CARRY, post-1.0 (parity-coverage)	ardopcf 1.0.4.1.2 added a built-in browser GUI (ardopcf -G 8514, then localhost:8514); the catalog carries 1.0.4.1.3	SUPERSEDE by ardopcf itself; the packet profile page documents the flag, as it does pat http
Morse Runner	Reserved to maintainer, conditional: "if Morse Runner CE builds, that is carried and Wine leaves 1.0"	Morse Runner Community Edition (w7sst/MorseRunner, Lazarus/Free Pascal) has a Linux port and 1.86 in preparation as of September 2026	Test the build on Debian 13 (Lazarus is apt: lazarus, fpc); carry on success under morse-training, which settles the conditional
chattervox	The one NEEDS-DECISION unit	Last release 0.7.0 (2020); README last touched 2019; prior-art.md already lists it dead	RETIRE, world-changed/abandoned, on a D-032 head-commit check — the same verdict prior-art.md reached, not yet applied to dispositions.md
tar1090	"documented gap rather than a shim" (readsb's manifest)	Unchanged upstream: a curl \| bash installer. But what it installs is static files plus a lighttpd or nginx stanza and a systemd unit	~~A tar1090 manifest from a tagged release, serving under lighttpd~~ Added 2026-10-02 as a data unit served by reference serve on 127.0.0.1, with no lighttpd: upstream has no tags, so the pin is a commit (see Measured since)
dream	REVIVE, blocked on libqt5webkit5-dev	Unchanged; the question in not-carried.md (is WebKit only the optional dashboard?) still needs the source read	No change
FoxTelem	Post-1.0 pending an AMSAT census	Unchanged	No change
VARA / VARIM	Post-1.0, Wine prefix	Mercury (below) is an open VARA-API-compatible modem released May 2026; the case for a Wine prefix is weaker	Re-rank after Mercury is measured

Measured since, 2026-09-30 — the three re-rulings. All three were applied, and two of this table's claims were wrong (dispositions.md has the evidence):

• ARDOPGUI is SUPERSEDE by ardopcf's web GUI, as recommended; the pin already serves it, so no bump.
• Morse Runner CE has no Linux port. At v1.86 (released 2026-09-25, no longer "in preparation") it is a Delphi tree that stops at Can't find unit Windows under Free Pascal, and its README lists a Linux port as a goal. The native build that exists is the older Lazarus port of 1.68 (zmetzing/MorseRunner); it built and ran on Debian 13, so under the conditional's second clause it is CARRY as morse-runner, tagged morse-training and cw, in no profile.
• chattervox's last release is 2019-03-17, not 2020: the 2020 date was GitHub's push field (D-032). RETIRE, as recommended.

A9. The guides the standard requires do not exist

CLAUDE.md names docs/guides/ as "task-oriented: digital modes, APRS, satellite, packet, SDR" and sets the bar at a licensed ham reaching a working digital-modes station from a fresh Parrot install without asking anyone. Four guides exist under docs/guides/ (conference operating, offline navigation, propagation, Rayhunter); the five named ones do not. The profile pages carry good manual_configuration prose but no walk-through. The order that serves the standard (each a new page under docs/guides/):

1. digital-modes — rig, audio (A3), time (A4), the callsign dialogs (A1), the first FT8 decode, the first QSO.
2. packet — Direwolf's file, KISS to pat, pat http, the first Winlink message; ARDOP with its web GUI.
3. aprs — Direwolf as the modem, Xastir or YAAC, the iGate decision (the Direwolf manifest already says why it is a decision).
4. satellite — gpredict, TLEs, Doppler over rigctld, a rotator over rotctld, SatDump for the images.
5. sdr — which receiver for which job (the overlaps.md verdicts, as a page), the first decode with each carried decoder.

Each guide is also where the A1 "type it once" instructions live for the programs that get a page rather than a block.

B. Per-category findings

Abbreviations in the route column: apt (a distribution package; unmeasured means no target sweep yet), source, git, binary, venv, data, apt_repos (D-040), each an existing backend. Every candidate needs a D-032 liveness check, a licence check (docs/reference/licence-verification.md's shape) and the archive sweep before a manifest. Nothing here is transmit-capable except where said.

Group 1 — Operate the Station

Category	Carried	Candidate	What it adds	Route	Recommendation
rig-control	12	piHPSDR (g0orx/pihpsdr, DL1YCF and G0ORX, GPL-3.0; issues active September 2026)	The operator program for OpenHPSDR protocol 1 and 2 radios: ANAN and the Hermes Lite 2, which has one of the largest homebrew-SDR communities there is. gr-hpsdr is carried and gives those owners a GNU Radio source and nothing to operate with	source (make)	ADD, rig-control + sdr-receivers; deskhpsdr (DL1BZ) is the desktop-oriented fork and the overlaps.md row
rig-control		rigctld as a service; ser2net	See A2	config + apt	A2
rig-control		Documentation gap: the maintainer's own FT-991A is not in CHIRP (chirpmyradio issues 6685, 7987, 9073: no developer with the radio) and is programmed by Yaesu's Windows-only ADMS-991A; the BTECH UV-50PRO has no CHIRP driver and is programmed only through BTECH's Bluetooth phone app	—	—	Say so on the chirp page's known_problems and on a hardware/ page for each radio; both radios are CAT-controllable through hamlib (1035 for the FT-991; this page first said 3073, which is the IC-7300 — corrected from rigctl -l, hamlib 4.7.2, on the field laptop, 2026-09-29), which is the part that matters for the station
logging	13	—	Coverage is complete against every source; the gap is offline lookup (A6)		No add
contest	5	so2sdr (N4OGW, GPL-3.0; apt unmeasured)	An SO2R-capable contest logger with a Linux-native GUI, the shape not1mm and tlf do not have	apt	Measure the sweep; ADD if in Debian 13
dx-cluster	2	—	The Reverse Beacon Network is a cluster feed and needs a page, not a unit		A sentence in the DX-cluster pages
time-frequency	3	chrony	See A4	apt + config	A4
gps-gnss	7	PyGPSClient (PyPI pygpsclient, BSD-3; NMEA, UBX, RTCM3, NTRIP)	Receiver configuration with a GUI; the u-center replacement	venv	ADD
gps-gnss		gpsprune (apt, in trixie per the search result; measured 2026-09-30: all seven targets), gpxviewer (trixie), foxtrotgps (sid)	Track viewing and editing; gpsbabel converts but shows nothing	apt	~~ADD gpsprune~~ Not added: D-061 measured it and does not carry it (see Measured since); foxtrotgps is online-first and adds nothing over Navit
gps-gnss		rtklib (RTK and PPK post-processing; apt unmeasured)	Centimetre positioning for antenna surveys	apt	Post-1.0 on demand
navigation-maps	14	OpenCPN (Debian 13 1:5.10.2+dfsg-1 per the search result, confirmed by the 2026-09-30 sweep, which found it on all seven targets; GPL-2.0)	A marine chart plotter that takes position from gpsd and AIS targets from rtl-ais or ais-catcher over NMEA — the one program that makes the ships category more than a decoder. Charts are data: NOAA ENC and RNC are public domain and fetched by region, the D-057 shape again	apt now; charts as a data/regional unit later	ADD, tagged ships + navigation-maps (added 2026-09-30, navigation); charts a follow-on
navigation-maps		Organic Maps (Linux desktop: a Qt 6 source build the project says needs 20 GB and has no feature parity; Flathub only)	Turn-by-turn on the desktop with a modern UI	none acceptable today (Flatpak is a measured zero, D-014)	Not carried; revisit if a distribution packages it
navigation-maps		qgis, kiwix, dict, mbtileserver	Already waiting on ETC sub-project 5		No change
locators, dashboards		—			No add

Group 2 — Digital Modes & Morse

Category	Carried	Candidate	What it adds	Route	Recommendation
weak-signal	8	—	Coverage complete. wsprdaemon is a KiwiSDR/RX888 appliance, not a station program; not carried		No add
keyboard-modes, nbems	6, 5	—	Complete (linpsk is dead: last in buster)		No add
digital-voice	2	FreeDV 2.x with RADEV1	The catalog's freedv is apt 1.8.11; the neural RADE mode that made FreeDV competitive with SSB shipped in 2.0 (January 2025) and 2.2.x (2026), and is not in any target's archive. Upstream's Linux route is an AppImage (refused by name) or a build script	source (cmake, with the RADE dependencies)	ADD as a second block or a freedv-rade unit; the apt one stays for operators who want the packaged version
digital-voice		DroidStar (nostar/DroidStar, AD8DP, GPL-3.0; Linux .deb and AppImage from the 9M2PJU distribution)	DMR, D-STAR, YSF, P25, NXDN and M17 over reflectors with no radio; the built-in vocoders are the point. Gives the M17 roadmap (D-007) something to run today while mvoice is blocked	binary (.deb, pinned and hashed) or source (Qt)	ADD; state that reflector traffic is transmit in the network sense and needs a callsign registration, not a consent gate
digital-voice, sdr-receivers		QRadioLink (Codeberg, GPL-3.0; needs GNU Radio 3.10, Qt 5.15; upstream Debian 12 .debs only)	A full SDR transceiver with Codec2/Opus digital voice for Pluto, LimeSDR, USRP; transmit-capable	source (qmake)	ADD post-1.0, consent-gated on transmit (D-021)
sstv-atv	3	hamfax (apt; radiofax), slowrx (SSTV receive-only), leansdr (DATV receive; all unmeasured)	xwefax already covers fax; qsstv covers SSTV both ways	apt	Low; measure and ADD only where the sweep finds them
echolink-repeaters	6	—	Track B (D-046)		No change
cw, morse-training	17, 10	Morse Runner CE	See A8	source (Lazarus)	Test the build

Group 3 — Packet, Mesh & Emergency Comms

Category	Carried	Candidate	What it adds	Route	Recommendation
soundcard-modems, winlink	8, 4	Mercury (Rhizomatica, GPL-3.0, C; released May 2026; Debian binaries and source)	An open HF OFDM ARQ modem that speaks VARA's TCP API, so pat (which already supports VARA) uses it unchanged. Rhizomatica's own comparison has it level with VARA HF at good SNR and ahead in noise. The first free answer to the VARA gap this project deferred post-1.0	source	ADD to the packet core; it re-ranks VARA (A8). Transmit-capable in the ordinary station sense, like ardopcf
soundcard-modems, winlink		FreeDATA (DJ2LS/FreeDATA, GPL-3.0; v0.17.x, active July 2026; codec2 modems, server plus GUI client)	Messaging and file transfer over HF with its own protocol and station-to-station chat; not Winlink	source or venv (the server is Python)	ADD, winlink-adjacent under soundcard-modems; pair with Mercury in one overlaps.md row
soundcard-modems		soundmodem (Thomas Sailer; apt unmeasured)	The other soundcard modem; Direwolf superseded it in practice	apt	Not carried unless asked
packet-nodes, winlink	8, 4	rmsgw (nwdigitalradio/rmsgw, GPL with one binary-only library; last commit December 2024)	The Linux Winlink RMS gateway — the category summary promises "gateways" and none is carried. A gateway needs Winlink sysop authorisation and a password, which is a secret, not station data	source (autotools)	ADD post-1.0, with the note that the password is entered by the operator and never templated (A1)
packet-nodes, aprs		aprsc (APRS-IS server; apt unmeasured)	Running your own APRS-IS server; niche	apt	On demand
winlink		ARDOPGUI	See A8: ardopcf now has it built in		SUPERSEDE
mesh	2	Reticulum, MeshCore, meshtasticd, tncattach, PlatformIO	All in issue #105 (track C)		No change here
emcomm, weather-sensors	9, 4	EAS/SAME alerts from NOAA Weather Radio	multimon-ng (carried) decodes EAS; dsame3 (Python 3 fork of dsame) turns the header into readable text and can run a command. rtl_fm to multimon-ng -a EAS to dsame3 is the whole pipeline and it is exactly what an EMCOMM laptop should be doing in the background with the network down	venv for dsame3 (licence to check), plus a page	ADD the page first; the unit follows
weather-sensors		weewx (own signed apt repository; not in Debian, which has only indi-weewx-json)	Weather-station software for fifty station models, many of which rtl-433 already hears	apt_repos (D-040)	Post-1.0 on demand
email-clients	5	—			No add

Group 4 — SDR & Listening

Category	Carried	Candidate	What it adds	Route	Recommendation
sdr-receivers	10	OpenWebRX+ (luarvique; signed repository with Debian 13 packages)	A multi-user web receiver: share one SDR with the whole EMCOMM room, or with yourself on a phone. Also the missing half of the Skywave on-ramp — Skywave's users connect to servers like this one	apt_repos (D-040)	ADD post-1.0; the server-on-a-laptop case needs a page on what it exposes
sdr-receivers		ka9q-radio (Phil Karn; GPL; RTL-SDR, Airspy, Funcube, RX888, optionally Fobos, HydraSDR, HackRF, BladeRF)	Demodulate every channel in a band at once; radiosonde-auto-rx (carried) documents running on top of it	git (untagged? — D-024 check needed)	Post-1.0; measure the pin source first
sdr-receivers		kiwiclient	Already in skywave-inventory.md as the maintained KiwiSDR client	venv	ADD, cheap
sdr-hardware	33	HydraSDR RFOne (libhydrasdr 1.1.1, SoapyHydraSDR, hydrasdr_433; the 2025–26 Airspy successor; not in any archive)	A device the catalog has no entry for at all: hardware class entry, host library, Soapy module	source (cmake) ×2	ADD a device and two units when a unit of hardware is available to measure; maintainer_verified: false until then (D-027)
sdr-hardware		Fobos SDR (rigexpert/libfobos, SoapyFobosSDR; the vendor of antscope2, already carried)	Same shape as HydraSDR	source ×2	Same
sdr-hardware		libiio-utils, iio-oscilloscope (apt; libiio is in trixie per the search result; measured 2026-09-30: libiio-utils on all seven targets)	The PlutoSDR's own tools; the device page already says the Pluto is "reached through libiio" and nothing installs it	apt	ADD libiio-utils beside the Pluto device (added 2026-09-30)
sdr-toolkits	12	gr-fosphor (trixie 3.9~0.974ab2f-1+b15 per the search result; measured 2026-09-30: 3.9~git20240323.74d54fc-1+b10, all seven targets)	The GPU waterfall every GNU Radio demo uses; ~~a Blend sdr task member~~ not in the Blend's task files (blend-inventory.md), and not in the catalog	apt	ADD (added 2026-09-30)
sdr-toolkits		gr-dab, gr-ais, gr-iqbal (unmeasured)	Tier 3 governs the rest	apt where packaged	Measure; ADD only the packaged ones
signal-analysis	4	URH, SigDigger	DragonOS Tier 2 (SCOPE.md stage 7)		No change
signal-analysis, broadcast		redsea (RDS decoder, CLI; apt unmeasured)	The command-line RDS decoder beside gr-rds	apt	Measure; ADD if packaged
broadcast	7	nrsc5 (theori-io/nrsc5, GPL-3.0; cmake) with NRSC5 Studio (Rust GUI, June 2026, Linux)	HD Radio — the only digital broadcast standard in North America, and the catalog has DAB twice and HD Radio not at all	source	ADD nrsc5; the GUI once it has a tagged release
broadcast		Qt-DAB (JvanKatwijk; active 2026; source build by script)	A third DAB decoder beside welle-io and dablin	source	Not carried; two DAB decoders is enough (SCOPE.md: "four ADS-B decoders and no opinion")
aircraft	10	dump978-fa (FlightAware; C++ over SoapySDR; .deb built from source with debhelper)	978 MHz UAT, the second ADS-B link, used by US general aviation below 18,000 ft. The category has 1090 MHz three ways and 978 not at all	source (cmake)	ADD; US-only relevance stated on the page
aircraft		tar1090	See A8	git	ADD as described
aircraft		piaware / dump1090-fa (FlightAware apt repository), the adsb.fi and ADS-B Exchange feeders	Feeding a network is the operator's choice and each network has its own client; mlat-client-adsbfi is carried	apt_repos	A page listing the feeders, no more units
aircraft		JAERO, OGN (ogn-rf)	Tier 2; low		No change
ships	4	OpenCPN (measured 2026-09-30: all seven targets)	See Group 1	apt	ADD (added 2026-09-30, navigation)
ships		Gap: NAVTEX is decoded by fldigi (carried) — say so on the page. No open DSC decoder exists	—	—	A sentence; nothing to carry
pagers-decoders	3	—			No add

Group 5 — Satellites & Propagation

Category	Carried	Candidate	What it adds	Route	Recommendation
satellite-tracking	1	predict (KD2BD, GPL-2.0; apt, in testing per the search result — wrong: in no archive, measured 2026-09-30, see Measured since)	The one-member category. PREDICT is the terminal tracker gpredict grew from, runs headless with a network server other programs poll, and is what a Pi ground station uses	~~apt~~ none packaged	~~ADD~~ Not added: no target offers it
satellite-tracking		SatNOGS client (2.1.1, December 2025; pip; needs hamlib and gpsd Python bindings)	Turns the station into a ground station for the network — the open-source satellite community's own project	venv	ADD post-1.0 as an appliance unit; needs a page on what it uploads
satellite-tracking		rotctld configuration	libhamlib-utils carries it; nothing configures a rotator	config + page	Part of A2
satellite-decoding	3	goestools (pietern/goestools; GOES HRIT/LRIT)	SatDump (carried) decodes GOES; goestools' goesrecv is still the community's lock-and-signal-strength tool for aiming a dish	source (cmake)	Optional; a note on the SatDump page is enough until asked
propagation	7	Signal-Server (Cloud-RF, the multi-threaded SPLAT fork; carried by DragonOS)	Coverage maps in minutes instead of hours, from the same terrain (A5)	source (make)	ADD with the splat-sdf converter (added 2026-10-02: W3AXL's fork, CMake, in antenna; Cloud-RF's repository holds only a README now)

Group 6 — Antennas, Bench & Programming

Category	Carried	Candidate	What it adds	Route	Recommendation
antenna	10	necpp, linsmith, transcalc, antennavis, meep, openems	The Blend's electronics-radio-dev metapackage recommends antennavis, gsmc, meep, nec2c, openems, transcalc, xnec2c, yagiuda and suggests linsmith, qucs, xsmc-calc; the catalog carries that metapackage with install_recommends at the default, so the recommended ones already arrive without a manifest, a page or a menu entry	apt	Give each Recommends member its own manifest so the menu and the package reference see them; xnecview is dead (last in bullseye)
antenna-analysers	3	LibreVNA GUI (jankae/LibreVNA, GPL-3.0; 1.6.5; Qt; Ubuntu binary and source)	The open-hardware VNA that nanovna-saver does not speak to	source (qmake/cmake)	ADD
spectrum-analysers	2	—	hackrf_sweep is in hackrf; the RF Explorer's Python library is niche		No add
electronics	6	The other Blend electronics metapackages: electronics-pcb (KiCad), electronics-simulation (ngspice, Qucs-S), electronics-microcontrollers, and sigrok/pulseview	The catalog carries one Debian Electronics task; the bench needs three more, in the same metapackage shape	apt	ADD the metapackages, not the members
radio-programming	4	k5prog (sq5bpf/k5prog, GPL; C) and the UV-K5 CHIRP drivers (egzumer, armel/F4HWN)	The Quansheng UV-K5 is the most-modified handheld of the decade; chirp (carried) programs it only through a firmware-specific driver, and flashing custom firmware on Linux is k5prog	source (make); the CHIRP driver is a file the page points at	ADD k5prog; page on the driver
radio-programming, dmr		OpenRTX radio_tool	Flashes OpenRTX (M17-capable) onto TYT MD-UV380/390 and others from Linux; D-026 says carry the means of talking to the device	source (cmake)	ADD
radio-programming		editcp (AnyTone)	qdmr covers AnyTone; editcp is dormant		Not carried
programmer	8	stm32flash (trixie; measured 2026-09-30: 0.7 on all seven targets), teensy-loader-cli (trixie), dfu-programmer, bossa-cli, picotool, mspdebug (unmeasured)	STM32 serial bootloaders are how MMDVM boards, NanoVNAs and many radios are flashed; RP2040/RP2350 boards need picotool	apt	ADD stm32flash (added 2026-09-30) and picotool first; the class file is generated (gen_programmer_class.py reads the rules)
programmer		adafruit-nrfutil (PyPI) and a UF2 note	The maintainer's nRF52840, T-Echo and RAK nodes all flash by UF2 drag-and-drop or adafruit-nrfutil dfu serial; nothing in the catalog says so	venv + page	ADD, with the Meshtastic device page carrying the UF2 procedure
programmer		cc2538-bsl and catnip (Electronic Cats)	The CatSniffer V3's firmware loader; Electronic Cats' own docs prefer catnip, and Kismet's page says cc2538-bsl worked and catnip did not in their testing	venv (pyserial, intelhex)	ADD cc2538-bsl on the CatSniffer page; the device is owned, so this is measurable now
programmer		pyocd, platformio	pyocd for nRF and STM32 over SWD; PlatformIO is the Meshtastic/MeshCore build tool (#105)	venv	On demand
serial-terminals	5	ser2net (measured 2026-09-30: all seven targets)	See A2	apt	ADD (added 2026-09-30, in no profile: see Measured since)
device-support	8	qpwgraph (measured 2026-09-30: all seven targets), pavucontrol	See A3	apt	ADD (qpwgraph added 2026-09-30, digital-modes)

Group 7 — RF Security & Research

Out of this pass except: A7 (Kismet), the CatSniffer loader above, and one observation for the record — the maintainer's owned devices in this group (CatSniffer, Minino, Free-WiLi 2, Clip-Boy, C5 Wardriver) each have a hardware/ page, and none yet lists the software that talks to it beyond the flasher. That is a per-device page task under D-029, not a catalog gap.

Group 8 — Learn & Practise

Category	Carried	Candidate	Recommendation
exams	3	—	Complete for US and Canada; a UK (RSGB) or Australian source would be an ADD on request
morse-training	10	Morse Runner CE	A8
references	2	FCC ULS (A6); kiwix, dict (sub-project 5)	A6

Group 9 — Workstation

No gaps. usbip (sharing a USB SDR or rig to another machine over the network, in linux-tools) would be a page under troubleshooting, not a unit.

C. Thin categories, for the record

A category with one or two members is not wrong — the vocabulary is cut fine on purpose (D-055) — but each is worth a glance:

Category	Members	Note
satellite-tracking	1	SatNOGS above; predict is in no archive (2026-09-30)
dx-cluster	2	Complete
locators	2	Complete
mesh	2	Issue #105
digital-voice	2	FreeDV 2.x, DroidStar, QRadioLink above
spectrum-analysers	2	Complete for the carried hardware
references	2	A6
dmr	2	radio_tool above

D. Recommended order

Ranked by how many operators hit the gap on the first evening, against the cost:

1. A3 + A9.1 — the audio-routing guide and the digital-modes guide. Documentation only; unblocks the standard.
2. A1 — config_files for the eight plain-text units; direwolf first, because D-008 promised it by name.
3. A4 — chrony in station. One manifest, one page.
4. A7 — Kismet, and fix the stale prose.
5. Mercury and FreeDATA — the open HF-modem pair; re-rank VARA after.
6. predict, gr-fosphor, libiio-utils, stm32flash, ser2net, qpwgraph, gpsprune, OpenCPN — apt units, one PR each, after the sweep confirms them. (2026-09-30: swept, see Measured since; predict is in no archive and is not added; gpsprune is not added because D-061 rules it out.)
7. A2 — the rig as station data. A sub-project with a spec, like the navigation ones under docs/superpowers/specs/.
8. A5 — the splat-sdf converter and Signal-Server. (Built 2026-10-02, D-061's amendment of that date.)
9. A8 — the three re-rulings: ARDOPGUI, Morse Runner CE, Chattervox.
10. Source-built adds — piHPSDR, FreeDV 2.x, DroidStar, dump978-fa, nrsc5, LibreVNA, k5prog, radio_tool, tar1090; each a pinned build with a D-032 check first. (2026-10-01: piHPSDR and DroidStar carried, FreeDV 2.x not yet; 2026-10-02: tar1090 carried as a reference serve page; see Measured since.)
11. Hardware without hardware — HydraSDR and Fobos entries wait for a unit to measure with (D-027).

What needs the maintainer, and only the maintainer, is in Q-022.