BS
← All projectsBlockstream

Blockstream Jade

Firmware for Jade and Jade Plus Bitcoin hardware wallets.

BitcoinHardware walletsElevated watch
Repository coverage

557 commits in the local evidence base

Every captured commit receives deterministic security triage and a separate communication-quality score. Security candidates and broader second-pass signals receive full-patch Ollama analysis.

120security candidates80second-pass queue462AI analyses
0commits · 30 days
47commits · 60 days
209commits · 180 days
423commits · 365 days
Backfill bands
Aug 5 → Feb 6216 seen38 candidatesComplete
Feb 6 → Jun 6131 seen9 candidatesComplete
Jun 6 → Jul 657 seen10 candidatesComplete
Jul 6 → Aug 510 seen0 candidatesComplete
Commit communication

Does the history explain itself?

Message quality measures whether a commit identifies its scope, purpose, rationale, testing, and supporting references. It does not change the security-severity score.

57/100 average clarity
37Strong · 80–100
204Adequate · 60–79
275Thin · 40–59
41Opaque · 0–39
5security candidates with opaque commit messaging
Read the scoring rubric →
Developer activity

Who is changing the project?

Public Git author strings; identities are not independently verified.

DeveloperCommitsCandidatesAnalyzedHigh riskMessage avg.
Jon Griffiths32663289157
Mike Tolkachev541744153
Daniel Newton902078055
Mihailo Milenkovic878057
roderico633074
Valerio Vaccaro2226056
Jamie C. Driver525053
Gustavo Cateim917078
Austin-Fulbright414060
kkdao211068
quantpoet111072
cmiguelrn111045
Analysis record

Published AI watches

Last scanned 35 minutes ago

Moderate 57 AI analysisMessage 45 · Thin
BS BlockstreamBlockstream Jade BitcoinHardware wallets

process: whitelist reply message function calls

This commit adds a whitelist of approved functions that can be used when sending reply messages from the Jade hardware wallet to a connected computer or phone. Previously, any function could be passed in, which could allow a bug or malicio…

Function-pointer whitelist added to reply pathMultiple static reply helpers made global to enable whitelist comparisonAssertion tightened to validate callback before use
a59d14c2by Jon Griffiths+53−1513 files
No security note in commit
Low 42 AI analysisMessage 30 · Opaque
BS BlockstreamBlockstream Jade BitcoinHardware wallets

temporary_stack: whitelist temporary task function calls

This commit adds a whitelist to a special helper that lets certain functions run with extra memory stack space. Previously, any function could be passed to this helper. Now only a small list of approved functions is allowed, except during …

Addition of function-pointer whitelistRemoval of static linkage on internal functions to enable whitelist comparisonConditional bypass of whitelist only for debug/selfcheck builds
6a0fea51by Jon Griffiths+26−54 files
No security note in commit
Informational 15 AI analysisMessage 43 · Thin
BS BlockstreamBlockstream Jade BitcoinHardware wallets

consistency: fix typos

This commit is a large typo-fixing cleanup across 40 files. Almost every change corrects spelling mistakes in code comments, variable names, or user-facing text strings. There is no change to program logic, security behavior, or cryptograp…

251f7713by Jon Griffiths+96−9640 files
No security note in commit
Informational 15 AI analysisMessage 47 · Thin
BS BlockstreamBlockstream Jade BitcoinHardware wallets

docs: document get_bip85_bip39_entropy

This commit only updates documentation and removes a single comment line. It documents an existing feature that lets a user request encrypted BIP85-based BIP39 entropy from the Jade hardware wallet after on-device confirmation. There is no…

05a578adby Jon Griffiths+52−12 files
No security note in commit
Low 28 AI analysisMessage 50 · Thin
BS BlockstreamBlockstream Jade BitcoinHardware wallets

bip85: expose get_bip85_bip39_entropy over rpc in addition to qr

This commit moves a feature that derives BIP85 BIP39 entropy from being available only in debug builds to being available over the normal RPC interface. BIP85 lets a hardware wallet derive child seed phrases from a master seed. Making this…

Feature moved from debug-only to production RPC surfaceNo accompanying access-control or input-sanitization changes visible in diffBIP85 entropy derivation is a sensitive cryptographic operation
c6c5e860by Jon Griffiths+3−41 file
No security note in commit
Informational 19 AI analysisMessage 50 · Thin
BS BlockstreamBlockstream Jade BitcoinHardware wallets

versioninfo: add JADE_INTERFACE field to show which connection source holds the unlocked wallet

This commit adds a new read-only field, JADE_INTERFACE, to the device's version information reply. It simply tells a connected computer which communication path (USB, Bluetooth, QR mode, etc.) currently has control of the unlocked wallet. …

No authentication or authorization logic changedNo cryptographic or key-management code changedNo input parsing or memory-unsafe code added
2fe9df0bby Daniel Newton+66−56 files
No security note in commit
Low 44 AI analysisMessage 60 · Adequate
BS BlockstreamBlockstream Jade BitcoinHardware wallets

sign_tx/sign_psbt: extra check for BTC inputs

This commit adds overflow protection when adding up the total value of Bitcoin transaction inputs in Blockstream Jade's signing code. Previously, the firmware used a plain addition that could silently wrap around if a specially crafted tra…

Integer overflow hardening in financial-amount accumulationDefensive checked_add_uint64() introduced where plain addition existedTransaction signing path now rejects overflowed input totals
41d7e8b4by Daniel Newton+15−32 files
No security note in commit
Informational 19 AI analysisMessage 62 · Adequate
BS BlockstreamBlockstream Jade BitcoinHardware wallets

build: move NVS cache objects that can pin the DRAM heap to SPIRAM if available

This commit changes how the Jade hardware wallet's firmware uses memory. It moves some internal storage cache objects from the main RAM (DRAM) to external RAM (PSRAM/SPIRAM) when available. This is a memory-optimization change, not a direc…

Memory layout change: moves NVS cache from DRAM to SPIRAMNo direct security fix or vulnerability patch present in diffNo mention of CVE, security researcher, or security advisory in commit
329f73c5by Daniel Newton+7−17 files
No security note in commit
Informational 15 AI analysisMessage 60 · Adequate
BS BlockstreamBlockstream Jade BitcoinHardware wallets

tests: migrate pin action tests from test_jade to rpc tests

This commit simply moves existing test cases from one test file to another. It does not change any production code, security behavior, or fix any vulnerability. The tests check that the Jade device correctly rejects invalid parameters for …

be0e6b3dby Mike Tolkachev+42−402 files
No security note in commit
Informational 15 AI analysisMessage 60 · Adequate
BS BlockstreamBlockstream Jade BitcoinHardware wallets

tests: add tests for pin actions via debug_set_pin

This commit only adds a new test file that exercises PIN-related features of the Jade hardware wallet using a debug-only RPC command. It does not change any production code, introduce new functionality, or alter security behavior. The test…

dd3d5b22by Mike Tolkachev+334−01 file
No security note in commit
Moderate 62 AI analysisMessage 50 · Thin
BS BlockstreamBlockstream Jade BitcoinHardware wallets

process: require user to permit get_identity_shared_key RPC

This commit adds a user confirmation step before a hardware wallet will share an identity-derived shared secret. Previously, the get_identity_shared_key RPC could compute and return the shared key without the user pressing approve on the d…

RPC that derives/releases a cryptographic shared secret now requires explicit on-device user approvalInput validation added: identity string must be printable before being shown in a security promptNew dedicated accept/reject button event IDs separate this flow from sign_identity
7ca0d26cby Daniel Newton+70−116 files
No security note in commit
Informational 20 AI analysisMessage 60 · Adequate
BS BlockstreamBlockstream Jade BitcoinHardware wallets

auth: add unattended pin selection for rpc tests

This commit adds a debug-only feature that lets automated tests set or change the device's PIN over an internal RPC message. It only compiles when special debug flags are enabled and is not present in normal production firmware. The change…

New RPC endpoint debug_set_pin can set/change the unlock PINPIN-handling code path modified to read from a mutable debug context instead of a fixed constantCode is wrapped in CONFIG_DEBUG_UNATTENDED_CI and CONFIG_DEBUG_MODE preprocessor guards
309ee8bbby Mike Tolkachev+100−205 files
No security note in commit
Moderate 59 AI analysisMessage 65 · Adequate
BS BlockstreamBlockstream Jade BitcoinHardware wallets

pin: disallow pin update on initialized units in debug mode

This commit removes a special debug-only exception that previously allowed the PIN server's public key to be changed on a Jade hardware wallet that already had a wallet set up. In normal operation, changing this public key on an initialize…

Removal of debug-mode bypass for a security-critical authorization checkProtection against pinserver public key substitution on initialized devicesDefense-in-depth for wallet PIN/server trust binding
a5667efdby Jon Griffiths+1−31 file
Vendor flagged security relevance
Informational 15 AI analysisMessage 95 · Strong
BS BlockstreamBlockstream Jade BitcoinHardware wallets

tests: fix mnemonic handling following seed changes

This commit only changes test code for the Blockstream Jade hardware wallet. It fixes how tests set and reset the device mnemonic/seed so tests run correctly after earlier seed-handling changes. There is no change to the actual wallet firm…

85e87d4eby Jon Griffiths+32−377 files
No security note in commit
Low 32 AI analysisMessage 35 · Opaque
BS BlockstreamBlockstream Jade BitcoinHardware wallets

malloc_ext: fix JADE_CALLOC_DRAM

This commit fixes a macro definition for a memory-allocation helper. The macro was accidentally written to take only one argument ('size') even though the underlying function it calls needs two arguments ('number of items' and 'size of eac…

Memory-allocation wrapper macro arity mismatch fixedPotential for undefined behavior if mismatched macro were usedNo explicit security claim or CVE referenced in commit
828835ccby Daniel Newton+1−11 file
No security note in commit
Low 43 AI analysisMessage 68 · Adequate
BS BlockstreamBlockstream Jade BitcoinHardware wallets

sign_psbt: fix reply count for exact chunk sizes

This commit fixes a bug in how Blockstream Jade counts the number of message chunks needed to send back a signed Bitcoin transaction (PSBT). When the transaction size was an exact multiple of the chunk size, the device told the host there …

Integer division off-by-one in chunk/message countDevice abort on assertion when host requests non-existent chunkDenial-of-service-like symptom: valid signing flow can crash the device
4f1a1565by kkdao+38−22 files
No security note in commit
Low 37 AI analysisMessage 73 · Adequate
BS BlockstreamBlockstream Jade BitcoinHardware wallets

ui: allow tapping directly on touchscreen elements

This commit adds a new touchscreen mode for some Jade hardware wallets where users can tap on-screen buttons directly instead of only using the three virtual buttons below the screen. To reduce the risk of accidental or malicious confirmat…

New direct-touch input path added to GUI event handlingCritical-action buttons require long-press to mitigate stray or coerced tap confirmationsDestructive operations (reset, delete wallet, delete OTP) converted to critical yes/no prompts
a46aa30aby Bota+206−1111 files
Vendor flagged security relevance
Informational 13 AI analysisMessage 60 · Adequate
BS BlockstreamBlockstream Jade BitcoinHardware wallets

urldecode: use plain strlen in self-tests of libjade

This commit only changes the internal test file for a URL-decoding helper. It swaps a custom string-length macro for the standard C strlen() function inside test data. The actual product code that users rely on is not changed, and there is…

b39a7a20by Mike Tolkachev+27−291 file
No security note in commit
Moderate 50 AI analysisMessage 45 · Thin
BS BlockstreamBlockstream Jade BitcoinHardware wallets

assets: support tickers up to 24 characters

This commit updates the Blockstream Jade hardware wallet to support longer asset tickers (up to 24 characters) and adds validation to reject malformed tickers. It also fixes a potential display bug where a ticker longer than 8 characters c…

Buffer size increased from 8 to 25 bytes to match new ticker maximumNew input validation added for asset ticker length and character setsnprintf precision argument cast from size_t to int to avoid undefined behavior on some platforms
cc5859a6by Mike Tolkachev+36−53 files
No security note in commit
Low 29 AI analysisMessage 45 · Thin
BS BlockstreamBlockstream Jade BitcoinHardware wallets

assets: improve empty ticker support

This commit hardens how the Blockstream Jade hardware wallet handles assets that have no ticker symbol. Previously, if an asset's ticker was missing (NULL), the code could pass a NULL pointer to string-length functions, which can cause cra…

NULL-pointer dereference risk in asset ticker handlingDefensive null-check added before strlen()New test fixture for NULL ticker asset path
9d0d74d0by Mike Tolkachev+30−23 files
No security note in commit
Repository ledger

Explore captured commits

Expand any commit for its author, full message, clarity score, changed files, triage signals, analysis, and source link.

Lower-prioritydisplay: fix oob text renderingby Mike Tolkachev · 6eb42c24 · Sep 30, 2026 · 1 fileMessage 45 · ThinTriage 0Details
Commit message · Mike Tolkachev

display: fix oob text rendering

45/100 · ThinMessage clarity
✓ Descriptive subject✓ Names a concrete action or component! No meaningful explanatory body
Security candidatetests: add tests for pin actions via debug_set_pinby Mike Tolkachev · dd3d5b22 · Sep 29, 2026 · 1 fileMessage 60 · AdequateInformational 15Details
Commit message · Mike Tolkachev

tests: add tests for pin actions via debug_set_pin

60/100 · AdequateMessage clarity
✓ Specific, descriptive subject✓ Names a concrete action or component✓ Mentions testing or verification! No meaningful explanatory body
Why it was queued
access controlauthentication path
AI analysis · Informational 15/100

This commit only adds a new test file that exercises PIN-related features of the Jade hardware wallet using a debug-only RPC command. It does not change any production code, introduce new functionality, or alter security behavior. The tests verify existing PIN setup, unlock, retry-limit, and pinserver-configuration logic.

AI review queuedprocess: require user to permit get_identity_shared_key RPCby Daniel Newton · 7ca0d26c · Sep 29, 2026 · 6 filesMessage 50 · ThinModerate 62Details
Commit message · Daniel Newton

process: require user to permit get_identity_shared_key RPC

50/100 · ThinMessage clarity
✓ Specific, descriptive subject✓ Names a concrete action or component! No meaningful explanatory body
Why it was queued
signing or wallet pathsecond-pass: security-sensitive path
AI analysis · Moderate 62/100

This commit adds a user confirmation step before a hardware wallet will share an identity-derived shared secret. Previously, the get_identity_shared_key RPC could compute and return the shared key without the user pressing approve on the device. The change also rejects identity strings containing non-printable characters so the on-screen prompt cannot be tricked with hidden control characters. In short, the device now asks 'Share Identity Key?' and the user must tap accept before any secret is released.

Security candidateauth: add unattended pin selection for rpc testsby Mike Tolkachev · 309ee8bb · Sep 29, 2026 · 5 filesMessage 60 · AdequateInformational 20Details
Commit message · Mike Tolkachev

auth: add unattended pin selection for rpc tests

60/100 · AdequateMessage clarity
✓ Descriptive subject✓ Names a concrete action or component✓ Mentions testing or verification✓ Names security-relevant behavior explicitly! No meaningful explanatory body
Why it was queued
access controlauthentication path
AI analysis · Informational 20/100

This commit adds a debug-only feature that lets automated tests set or change the device's PIN over an internal RPC message. It only compiles when special debug flags are enabled and is not present in normal production firmware. The change itself is a test helper, but it touches the sensitive PIN-entry path, so it deserves careful review to ensure it cannot accidentally end up in a shipping build.

Security candidatepin: disallow pin update on initialized units in debug modeby Jon Griffiths · a5667efd · Sep 28, 2026 · 1 fileMessage 65 · AdequateModerate 59Details
Commit message · Jon Griffiths

pin: disallow pin update on initialized units in debug mode

The caller can debug_reset to work around this at any point.

65/100 · AdequateMessage clarity
✓ Specific, descriptive subject✓ Names a concrete action or component✓ Provides an explanatory body
Why it was queued
access controlboot or update pathauthentication path
AI analysis · Moderate 59/100

This commit removes a special debug-only exception that previously allowed the PIN server's public key to be changed on a Jade hardware wallet that already had a wallet set up. In normal operation, changing this public key on an initialized device is blocked to prevent an attacker from redirecting the device to a malicious PIN server and potentially unlocking or controlling the wallet. The debug exception created a security gap: if a debug-mode device (or an attacker able to trigger debug mode) was initialized, they could swap the PIN server key. The fix closes that gap by applying the same restriction regardless of debug mode. A legitimate developer who needs to change the key can still use debug_reset first.

Lower-prioritytests: remove seed special case handlingby Jon Griffiths · a8663f92 · Sep 28, 2026 · 5 filesMessage 78 · AdequateTriage 0Details
Commit message · Jon Griffiths

tests: remove seed special case handling

Use the existing logic for mnemonic handling which ensures tests are
ordered and seeds cached across test calls.

Note that using seeds for test cases is deprecated and should be removed
in the future, since they don't play well with e.g. descriptors.

78/100 · AdequateMessage clarity
✓ Descriptive subject✓ Names a concrete action or component✓ Provides detailed explanatory context✓ Mentions testing or verification
Security candidatetests: fix mnemonic handling following seed changesby Jon Griffiths · 85e87d4e · Sep 28, 2026 · 7 filesMessage 95 · StrongInformational 15Details
Commit message · Jon Griffiths

tests: fix mnemonic handling following seed changes

- Fix calling set_mnemonic directly instead of through the caching
config wrapper
- Rename invalidatecache to reset since these tests require device reset
- Handle reset explicitly and don't use a valid mnemonic for it, to
prevent any test becoming dependent on it and avoid the overhead of
setting the unused mnemonic
- Add the missing mnemonic argument to the tests, to make the reset mnemonic
actually reset the device (as the fixture is not executed otherwise)

95/100 · StrongMessage clarity
✓ Specific, descriptive subject✓ Names a concrete action or component✓ Provides detailed explanatory context✓ Explains rationale or failure mode✓ Mentions testing or verification
Why it was queued
defensive validationauthentication path
AI analysis · Informational 15/100

This commit only changes test code for the Blockstream Jade hardware wallet. It fixes how tests set and reset the device mnemonic/seed so tests run correctly after earlier seed-handling changes. There is no change to the actual wallet firmware or any user-facing security behavior.

Lower-priorityinput: add a dead zone between the touch navbar's prev/click/next zonesby roderico · 72ec8815 · Sep 27, 2026 · 1 fileMessage 83 · StrongTriage 0Details
Commit message · roderico

input: add a dead zone between the touch navbar's prev/click/next zones

The three touch zones met at exact boundaries, so a touch that landed
close to one - from imprecise aim - snapped to whichever side it happened
to fall on, which read as the touch panel being imprecise.

Add a small dead zone (CONFIG_DISPLAY_WIDTH / 24 either side) around both
boundaries. A touch that starts in it does nothing. A touch that started
in a zone and drifts into it is not cancelled - only reaching another zone
is - so it also gives a drifting finger a little slack before it counts as
having left its zone.

With CONFIG_DISPLAY_TOUCH_DIRECT, a touch that starts in the dead zone is
still a navbar press, so it does not reach the gui either.

Tested on real Waveshare Touch LCD 2 hardware, with
CONFIG_DISPLAY_TOUCH_DIRECT on and off.

83/100 · StrongMessage clarity
✓ Specific, descriptive subject✓ Names a concrete action or component✓ Provides detailed explanatory context✓ Mentions testing or verification
Lower-priorityinput: act on touch navbar release, and repeat prev/next while heldby roderico · ec1c515c · Sep 27, 2026 · 1 fileMessage 95 · StrongTriage 0Details
Commit message · roderico

input: act on touch navbar release, and repeat prev/next while held

The touch navbar fired prev/select/next as soon as a touch was detected,
and relied on a fixed 100ms delay to avoid re-firing while the finger was
still down. A touch held for longer than that fired the action again,
which read as the touch panel being overly sensitive. The zone was also
re-evaluated on every poll, so a finger drifting during a touch triggered
whichever zone it happened to be over.

Handle the navbar like the physical buttons (navbtns.inc) instead: a touch
belongs to the zone it started in and acts once, on release. The finger
leaving the navbar or reaching another zone cancels the touch, so drifting
can't trigger a neighbouring action - in particular an unintended select.

Holding prev/next keeps auto-repeating, which things like the keyboard
rely on for fast scrolling, but now with the physical buttons' long-press
timings (CONFIG_BUTTON_LONG_PRESS_TIME_MS and
CONFIG_BUTTON_LONG_PRESS_HOLD_SERIAL_TIME_MS), measured with the tick
count rather than by counting polls. Once it has repeated, nothing more
fires on release. Select never repeats.

With CONFIG_DISPLAY_TOUCH_DIRECT, presses that start on the navbar are
still kept from the gui, and any other press still goes to it, even if
dragged onto the navbar. Whether a press started on the navbar is tracked
apart from its zone, as a cancelled navbar press no longer has a zone but
must still not reach the gui.

A failed touch read is skipped, as before, rather than taken as the finger
being lifted, which would act mid-press and then again on the real
release.

This affects every board using the touchscreen input: the Waveshare Touch
LCD 2 and 3.5, the M5Stack Core 2 and Core S3, and the TTGO T-Watch S3.

Tested on real Waveshare Touch LCD 2 hardware, with
CONFIG_DISPLAY_TOUCH_DIRECT on and off.

95/100 · StrongMessage clarity
✓ Specific, descriptive subject✓ Names a concrete action or component✓ Provides detailed explanatory context✓ Explains rationale or failure mode✓ Mentions testing or verification
Lower-prioritycamera: show the camera preview at full width on portrait displaysby roderico · 169d0971 · Sep 27, 2026 · 1 fileMessage 83 · StrongTriage 0Details
Commit message · roderico

camera: show the camera preview at full width on portrait displays

The preview was limited to the central 70% of the screen width. On a
portrait display, such as the Waveshare Touch LCD 2, that is the tightest
constraint, so the preview showed noticeably less of the QR code than the
sensor was capturing - pushing the QR code farther from the camera to make
it "fit" the smaller preview. That does not help the scan: QR decoding uses
a fixed central crop of the sensor image, regardless of the preview size.

Nothing else occupies that side margin - the header buttons, label and any
progress bar are all drawn over the picture, not beside it - so on portrait
displays let the image use the full screen width. Landscape displays are
unchanged.

That width also feeds the scale factor, though, so keep choosing the
scale as if the image were limited to 70%. Widening the image then does
not also zoom in.

Sensor area shown in the preview on the Waveshare Touch LCD 2:
168x240 -> 240x240 (1:1 scale, unchanged).

Tested on real Waveshare Touch LCD 2 hardware. Checked with the libjade
GUI that the Jade v1 and v2 camera screens are unchanged.

83/100 · StrongMessage clarity
✓ Specific, descriptive subject✓ Names a concrete action or component✓ Provides detailed explanatory context✓ Mentions testing or verification
Lower-prioritystorage: warm up NVS partitions at init to reduce heap fragmentationby Daniel Newton · 80b50c19 · Sep 25, 2026 · 2 filesMessage 50 · ThinTriage 0Details
Commit message · Daniel Newton

storage: warm up NVS partitions at init to reduce heap fragmentation

50/100 · ThinMessage clarity
✓ Specific, descriptive subject✓ Names a concrete action or component! No meaningful explanatory body
Lower-priorityinput: skip button A when it is not configuredby kkdao · 6b3ed155 · Sep 25, 2026 · 1 fileMessage 68 · AdequateTriage 0Details
Commit message · kkdao

input: skip button A when it is not configured

One-button boards such as the ESP32-CAM and T-Camera Plus configure
only button B and leave button A at -1. The button driver rejects an
invalid GPIO, so creating button A failed its ESP_ERROR_CHECK and the
device aborted during startup, before button B was set up.

Only create button A and register its callbacks when it is configured,
as select_init() already does for the optional front and wheel
switches.

68/100 · AdequateMessage clarity
✓ Descriptive subject✓ Names a concrete action or component✓ Provides detailed explanatory context
Lower-prioritygui: keep the QR guide clear of the camera chrome on portrait displaysby oroderico · 0ba8e1b8 · Sep 25, 2026 · 1 fileMessage 95 · StrongTriage 0Details
Commit message · oroderico

gui: keep the QR guide clear of the camera chrome on portrait displays

The QR frame guide is sized from the whole camera activity area. On
landscape displays it spans the full height, with the exit and help
buttons and any progress bar drawn over it. On a portrait display, such
as the Waveshare Touch LCD 2, the square is limited by the width instead,
so its corner marks collide with the buttons in the top band and the
progress bar in the bottom one, and reach past the sides of the camera
preview image.

On portrait displays, size the guide from the area those leave free
instead. The bottom band is reserved, and the top band as well when the
resulting square would otherwise reach the header buttons. Landscape
displays are unchanged.

The band and button sizes are read from the vsplit/hsplit that
make_camera_activity() places under the guide node, using the same
node_get_split_data()/get_step() helpers the split renderers use, rather
than duplicating their percentages here - so the guide cannot drift out
of sync if that layout changes.

Tested on real Waveshare Touch LCD 2 hardware. Checked with the libjade
GUI that the Jade v1 and v2 camera screens are unchanged.

95/100 · StrongMessage clarity
✓ Specific, descriptive subject✓ Names a concrete action or component✓ Provides detailed explanatory context✓ Explains rationale or failure mode✓ Mentions testing or verification
AI review queuedmalloc_ext: fix JADE_CALLOC_DRAMby Daniel Newton · 828835cc · Sep 24, 2026 · 1 fileMessage 35 · OpaqueLow 32Details
Commit message · Daniel Newton

malloc_ext: fix JADE_CALLOC_DRAM

35/100 · OpaqueMessage clarity
✓ Descriptive subject! No meaningful explanatory body
Why it was queued
second-pass: opaque commit message
AI analysis · Low 32/100

This commit fixes a macro definition for a memory-allocation helper. The macro was accidentally written to take only one argument ('size') even though the underlying function it calls needs two arguments ('number of items' and 'size of each item'). Any code using this macro would fail to compile, or if it somehow compiled, would pass wrong arguments. The fix makes the macro accept both required arguments, matching the function it wraps. There is no direct evidence in the commit of a security vulnerability being exploited.

Lower-priorityui: tap the home menu with direct touchby Bota · 4c7374a3 · Sep 24, 2026 · 3 filesMessage 68 · AdequateTriage 0Details
Commit message · Bota

ui: tap the home menu with direct touch

The home menu has no buttons to tap, so with DISPLAY_TOUCH_DIRECT it
only worked with the virtual buttons. Now a tap on the big item on the
left opens it, and a tap on the small one on the right moves to that
item.

68/100 · AdequateMessage clarity
✓ Descriptive subject✓ Names a concrete action or component✓ Provides detailed explanatory context
Lower-priorityui: tap the carousel bar with direct touchby Bota · 59ee8956 · Sep 24, 2026 · 3 filesMessage 68 · AdequateTriage 0Details
Commit message · Bota

ui: tap the carousel bar with direct touch

Carousels have no buttons to tap, so with DISPLAY_TOUCH_DIRECT they only
worked with the virtual buttons. Now a tap on the arrows at the ends of
the bar moves back or forward, and a tap on the item in the middle
selects it.

68/100 · AdequateMessage clarity
✓ Descriptive subject✓ Names a concrete action or component✓ Provides detailed explanatory context
Security candidatesign_psbt: fix reply count for exact chunk sizesby kkdao · 4f1a1565 · Sep 24, 2026 · 2 filesMessage 68 · AdequateLow 43Details
Commit message · kkdao

sign_psbt: fix reply count for exact chunk sizes

When the signed PSBT length is an exact multiple of the output chunk
size, the reply advertised one message too many. The host then asked
for the extra chunk and the device aborted on an assertion. Count the
messages by rounding up instead.

68/100 · AdequateMessage clarity
✓ Descriptive subject✓ Names a concrete action or component✓ Provides detailed explanatory context
Why it was queued
signing boundarysigning or wallet path
AI analysis · Low 43/100

This commit fixes a bug in how Blockstream Jade counts the number of message chunks needed to send back a signed Bitcoin transaction (PSBT). When the transaction size was an exact multiple of the chunk size, the device told the host there was one extra chunk. When the host asked for that non-existent chunk, the device hit an internal assertion and aborted. The fix rounds the chunk count up correctly. A new test exercises sizes right at the chunk-size boundary.

AI review queuedui: allow tapping directly on touchscreen elementsby Bota · a46aa30a · Sep 24, 2026 · 11 filesMessage 73 · AdequateLow 37Details
Commit message · Bota

ui: allow tapping directly on touchscreen elements

With DISPLAY_TOUCH_DIRECT, on by default for the Waveshare touch boards,
a tap on a button selects and activates it. The three virtual buttons
below the screen still work as before, and screens without buttons are
still used with them.

Buttons that sign, install firmware or delete data need a long press, so
a stray tap cannot confirm them.

73/100 · AdequateMessage clarity
✓ Specific, descriptive subject✓ Names a concrete action or component✓ Provides detailed explanatory context
Why it was queued
signing or wallet pathsecond-pass: security-sensitive path
AI analysis · Low 37/100

This commit adds a new touchscreen mode for some Jade hardware wallets where users can tap on-screen buttons directly instead of only using the three virtual buttons below the screen. To reduce the risk of accidental or malicious confirmation, buttons that perform dangerous actions—such as signing transactions, installing firmware, or deleting wallet data—now require the user to press and hold for about 0.8 seconds. The change is a defensive usability/hardening feature rather than a fix for an already exploited bug.

Lower-priorityqemu: disable watchdog timersby Daniel Newton · 0980992c · Sep 24, 2026 · 2 filesMessage 62 · AdequateTriage 0Details
Commit message · Daniel Newton

qemu: disable watchdog timers

The `idf.py qemu` command also disables watchdog timers as they can be triggered due to non realtime emulation

62/100 · AdequateMessage clarity
✓ Descriptive subject✓ Provides an explanatory body✓ Explains rationale or failure mode
AI review queuedurldecode: use plain strlen in self-tests of libjadeby Mike Tolkachev · b39a7a20 · Sep 24, 2026 · 1 fileMessage 60 · AdequateInformational 13Details
Commit message · Mike Tolkachev

urldecode: use plain strlen in self-tests of libjade

60/100 · AdequateMessage clarity
✓ Specific, descriptive subject✓ Names a concrete action or component✓ Mentions testing or verification! No meaningful explanatory body
Why it was queued
parser or protocol pathsecond-pass: security-sensitive path
AI analysis · Informational 13/100

This commit only changes the internal test file for a URL-decoding helper. It swaps a custom string-length macro for the standard C strlen() function inside test data. The actual product code that users rely on is not changed, and there is no fix for a security flaw here.

AI review queuedassets: support tickers up to 24 charactersby Mike Tolkachev · cc5859a6 · Sep 24, 2026 · 3 filesMessage 45 · ThinModerate 50Details
Commit message · Mike Tolkachev

assets: support tickers up to 24 characters

45/100 · ThinMessage clarity
✓ Descriptive subject✓ Names a concrete action or component! No meaningful explanatory body
Why it was queued
signing or wallet pathsecond-pass: security-sensitive path
AI analysis · Moderate 50/100

This commit updates the Blockstream Jade hardware wallet to support longer asset tickers (up to 24 characters) and adds validation to reject malformed tickers. It also fixes a potential display bug where a ticker longer than 8 characters could overflow a small on-screen buffer, which might have caused the device to show incorrect information during transaction confirmation.

Lower-priorityassets: add tests for libjadeby Mike Tolkachev · 5c16bacd · Sep 24, 2026 · 1 fileMessage 55 · ThinTriage 0Details
Commit message · Mike Tolkachev

assets: add tests for libjade

55/100 · ThinMessage clarity
✓ Descriptive subject✓ Names a concrete action or component✓ Mentions testing or verification! No meaningful explanatory body
Lower-prioritygui: hold the gui mutex while flipping the displayby Bota · 05582cf1 · Sep 23, 2026 · 1 fileMessage 73 · AdequateTriage 0Details
Commit message · Bota

gui: hold the gui mutex while flipping the display

gui_set_flipped_orientation() talks to the panel from the calling task,
while the gui task may be flushing a frame at the same time.

The gui task holds the gui mutex while it flushes, so take it here too.

73/100 · AdequateMessage clarity
✓ Specific, descriptive subject✓ Names a concrete action or component✓ Provides detailed explanatory context
AI review queuedassets: improve empty ticker supportby Mike Tolkachev · 9d0d74d0 · Sep 23, 2026 · 3 filesMessage 45 · ThinLow 29Details
Commit message · Mike Tolkachev

assets: improve empty ticker support

45/100 · ThinMessage clarity
✓ Descriptive subject✓ Names a concrete action or component! No meaningful explanatory body
Why it was queued
signing or wallet pathsecond-pass: security-sensitive path
AI analysis · Low 29/100

This commit hardens how the Blockstream Jade hardware wallet handles assets that have no ticker symbol. Previously, if an asset's ticker was missing (NULL), the code could pass a NULL pointer to string-length functions, which can cause crashes or undefined behavior. The fix substitutes a single-space placeholder string for missing tickers and adds a test case for such an asset. It is a defensive robustness improvement rather than a confirmed exploitable vulnerability.

AI review queuedsign_tx: reject liquid txs containing inputs missing commitmentsby Daniel Newton · 29220b50 · Sep 22, 2026 · 1 fileMessage 50 · ThinModerate 59Details
Commit message · Daniel Newton

sign_tx: reject liquid txs containing inputs missing commitments

50/100 · ThinMessage clarity
✓ Specific, descriptive subject✓ Names a concrete action or component! No meaningful explanatory body
Why it was queued
signing or wallet pathsecond-pass: security-sensitive path
AI analysis · Moderate 59/100

This update fixes a flaw in Blockstream Jade's signing process for Liquid network transactions. Previously, if a transaction input lacked commitment data (the information needed to verify asset amounts and types), the device could still ask the user to approve an incomplete or misleading on-screen summary. The fix now rejects such transactions outright, preventing the user from being shown an unverified summary and potentially signing something they did not fully understand.