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.
Message quality measures whether a commit identifies its scope, purpose, rationale, testing, and supporting references. It does not change the security-severity score.
This commit removes the SHA-1 hashing function from the regular Trezor firmware. SHA-1 is an old, weak hash algorithm that is no longer considered secure for sensitive uses. The change deletes the code that exposes SHA-1 to apps running on…
Removal of a deprecated cryptographic primitive (SHA-1) from the firmware API surfaceReduction of attack surface and prevention of future misuse of a collision-vulnerable hashNo direct vulnerability patch or memory-safety bug is present in the diff
This commit changes how a cryptographic library sets up secure connections. Previously, the code automatically calculated the public key from the private key. Now, the caller must provide the public key directly. This is a code-quality ref…
Cryptographic key handling changed: public key is now supplied rather than derivedPotential reduction of key-mismatch risk if caller provides correct public keyNew null-pointer check added for static_public_key
This commit simply reorders the arguments of an internal function called dh() and updates every place that calls it. The actual math and security behavior are unchanged; it is a code cleanup with no security effect.
This commit simply renames two groups of internal status labels (called enums) in the Trezor firmware's cryptographic code. The old names like WAITING_FOR_REQUEST1 were shared between two different parts of the code, so the developer gave …
This commit fixes three places in the Trezor firmware's cryptographic code where a memory-clearing function could be called with a NULL pointer. In practice, passing NULL to memzero is harmless on Trezor's platform (it does nothing), but i…
NULL pointer passed to memory-zeroing helper in cryptographic codeUndefined behavior in C standard library contractDefensive hardening in Noise protocol implementation
This commit is a code cleanup (refactor) for the cryptographic handshake code used in Trezor devices. It changes how the other party's long-term public key is returned to the caller: instead of storing it inside an internal state structure…
Removal of long-term public key storage from internal handshake stateCaller-supplied output buffer for remote static public key reduces internal secret retentionError-path memzero of returned key material on failure
This commit is a build-system refactoring for Trezor firmware. It moves the mapping of command-line build options to Rust/cargo features out of hard-coded Rust logic and into per-project TOML files. It also adds a small change so debug bui…
Build-option mapping now lives in project.toml files, increasing the attack surface for supply-chain/build-configuration tamperingMakefile change automatically enables debug-link for PYOPT=0 debug buildsValidation added to reject storage_insecure_testing_mode in production builds
This commit only updates test data. It corrects the capitalization (checksum) of Ethereum addresses used in automated tests and refreshes the expected screen-shot hashes those tests compare against. There are no changes to the actual Trezo…
This commit changes the Trezor Python library so that, when setting up a Trezor Model One (the original Trezor 1 device), it now performs an entropy check by default if the device runs firmware 1.13.1 or newer. Previously, the library only…
Enables a previously disabled security/validation feature (entropy check) for a specific device modelAdds version-gated behavior to avoid errors on older firmwareDefensive hardening of wallet setup randomness verification
This commit is a routine cleanup that removes the old SCons build system files from the Trezor Core firmware repository. It deletes Makefiles, SConscripts, and related Python helper scripts, but does not change any firmware source code, cr…
This commit updates Trezor's embedded MicroPython interpreter from an older version to 1.28.0. It pulls in several upstream MicroPython bug fixes, including stricter buffer-size checks for converting integers to bytes, a new stack-safety A…
Synchronizes upstream MicroPython fixes that include buffer-size and stack-safety hardeningPrevents Ctrl+C interruption of frozen boot code, reducing denial-of-service/control-flow risk during bootFixes sys.stdout.buffer.write() return value, which could affect code relying on correct I/O semantics
This commit is a simple code cleanup that replaces a common MicroPython string-creation pattern with a new helper function. It does not change what the code does, only how it is written. There is no security issue visible in the change.
This is a large but straightforward internal cleanup: the project switched from using the MicroPython `ubinascii` module's `hexlify`/`unhexlify` functions to the standard Python `bytes.hex()` and `bytes.fromhex()` methods. The change remov…
This commit is a routine code cleanup in the Trezor firmware's embedded MicroPython modules. It replaces an older, two-step object allocation pattern with a newer helper that allocates memory and sets the object type in one step. The chang…
No security-relevant logic changes observedNo input validation changesNo memory safety bug fixes (e.g., no overflow, use-after-free, or uninitialized memory fixes)
This commit is a routine maintenance refactor. MicroPython removed its built-in 'utimeq' module, so Trezor copied the same code into its own firmware tree and renamed it to 'timeq'. There is no security bug being fixed and no new vulnerabi…
This is a code cleanup change in Trezor's firmware that swaps one MicroPython internal helper for another. It replaces calls that create byte or string objects from a vstr buffer with newer, purpose-built helpers. The commit message says t…
Refactor only: helper function renames with equivalent semanticsNew str helper adds UTF-8 validation; bytes helper does not validateNo input validation, length, or error-handling changes observed
This commit is a routine internal cleanup that switches how Trezor's firmware defines built-in MicroPython object types. It replaces older, manually-written type structures with a newer macro provided by the upstream MicroPython project. T…
This commit is a straightforward internal code refactor. It adds a small C wrapper file so that Rust code can call two specific internal MicroPython functions that are declared as static/inline in headers. There is no user-facing change, n…
This commit is a routine build-system update for the Trezor firmware. It changes where the build process looks for the MicroPython cross-compiler binary (mpy-cross) and synchronizes a configuration header with newer MicroPython code. There…
This commit updates the Trezor firmware build system to use a newer MicroPython mechanism for tracking special memory pointers called 'root pointers.' It removes hard-coded lists of these pointers from configuration files and instead colle…
GC root pointer handling changedBuild system now auto-collects root pointers instead of hard-coding themRemoves manual MICROPY_PORT_ROOT_POINTERS lists
chore(core): add testing tropic attestation key and certificate
[no changelog]
72/100 · AdequateMessage clarity
✓ Specific, descriptive subject✓ Names a concrete action or component✓ Uses a recognizable type or scope✓ Mentions testing or verification! No meaningful explanatory body
Why it was queued
update trust
AI analysis · Informational 18/100
This commit updates Trezor's internal test-only files that simulate a TROPIC01 secure chip. It replaces old self-signed test certificates with a new four-level test certificate chain and embeds keys directly into config files instead of referencing separate PEM files. The change is labeled as a testing chore, touches only test/development artifacts, and does not alter production firmware code. The most notable security-relevant detail is that a hardcoded X25519 private key is included in test data, but it is explicitly a test key already present in the public source tree and used only for the emulator/unix test build.
fix(core): disable writing MCU attestation certificate in firmware
[no changelog]
62/100 · AdequateMessage clarity
✓ Specific, descriptive subject✓ Names a concrete action or component✓ Uses a recognizable type or scope! No meaningful explanatory body
Why it was queued
update trust
AI analysis · Low 33/100
This commit changes a Trezor firmware function so that, outside of a special factory testing mode (PRODTEST), it refuses to write an MCU attestation certificate. Previously the function accepted and stored such a certificate. The change prevents firmware from installing or overwriting this certificate, likely to reduce attack surface, but the commit message gives no details about any specific security issue.
✓ Specific, descriptive subject✓ Names a concrete action or component✓ Uses a recognizable type or scope! No meaningful explanatory body
Why it was queued
update trust
AI analysis · Informational 24/100
This commit changes a factory-production testing tool (prodtest) so that the commands for writing and reading the device's MCU attestation certificate also work in the software emulator, instead of returning 'not implemented'. In the emulator, the written certificate is only stored in memory and disappears after reboot. The change removes compile-time guards that previously disabled these commands for emulator builds. There is no direct evidence in the commit that this fixes a security vulnerability; it appears to be a feature-enablement change for testing/development workflows.
refactor(core): move `compact_size()` from `common.py` to `tools.py`
[no changelog]
62/100 · AdequateMessage clarity
✓ Specific, descriptive subject✓ Names a concrete action or component✓ Uses a recognizable type or scope! No meaningful explanatory body
Why it was queued
signing or wallet pathauthentication path
AI analysis · Informational 15/100
This commit simply moves a helper function called compact_size from one internal file to another and updates the import statements in files that use it. There is no change to what the function does, no bug fix, and no security impact.
✓ Descriptive subject✓ Uses a recognizable type or scope✓ Mentions testing or verification! No meaningful explanatory body
Why it was queued
authentication path
AI analysis · Informational 15/100
This commit only adds a new automated test that checks whether Trezor hardware devices can prove their identity using an MCU-based digital signature. It does not change any firmware code that runs on the device, nor does it fix or introduce any security vulnerability. It is a testing-only change.
fix(core/python): fix encoding in authentication challenge
[no changelog]
67/100 · AdequateMessage clarity
✓ Specific, descriptive subject✓ Names a concrete action or component✓ Uses a recognizable type or scope✓ Names security-relevant behavior explicitly! No meaningful explanatory body
Why it was queued
access controlauthentication path
AI analysis · Low 44/100
This commit fixes how the Trezor Python library builds an authentication challenge. Previously, the code encoded the length of a header and the length of the challenge as a single byte, which only works for values up to 255. The fix uses a variable-length 'compact size' encoding instead, which can handle larger values. If a challenge or header were ever 256 bytes or longer, the old code would produce an incorrect byte string, which could cause authentication to fail or behave unexpectedly. There is no direct evidence in the commit that this was exploitable as a security vulnerability, but it is a correctness fix in a security-sensitive authentication path.
✓ Descriptive subject✓ Uses a recognizable type or scope✓ Mentions testing or verification! No meaningful explanatory body
Why it was queued
authentication path
AI analysis · Informational 15/100
This is a minor code cleanup in the test suite. It replaces a manually constructed challenge message with a helper function that does the same thing. There is no change to the actual Trezor firmware or to how device authentication works in production.
refactor(core/python): factor out `get_challenge_message()`
[no changelog]
62/100 · AdequateMessage clarity
✓ Specific, descriptive subject✓ Names a concrete action or component✓ Uses a recognizable type or scope! No meaningful explanatory body
Why it was queued
authentication path
AI analysis · Informational 15/100
This commit is a simple code cleanup: it moves a small block of code that builds an authentication challenge message into its own reusable function. There is no change to behavior, no bug fix, and no security-related change.
Move `if __debug__` address/signature asserts out of production modules (`ethereum/yielding_vaults.py`, `ethereum/yielding.py`, `ethereum/clear_signing_definitions.py`, `tron/consts.py`) into unit tests under `core/tests`, so they no longer inflate debug build firmwares.
- `test_apps.ethereum.yielding_vaults.py`: asserts hardcoded ETH vault and token address bytes match their published Etherscan hex values - `test_apps.ethereum.yielding.py`: asserts ERC-4626 / claim function selectors match keccak256 of their signatures, and Merkl distributor address matches its hex literal - `test_apps.ethereum.clear_signing_definitions.py`: asserts approve/transfer function selectors match keccak256 of their signatures - `test_apps.tron.consts.py`: asserts hardcoded TRC-20 token address bytes match their published base58check addresses
Also fix a typo in USDT link comment.
[no changelog]
95/100 · StrongMessage clarity
✓ Descriptive subject✓ Names a concrete action or component✓ Uses a recognizable type or scope✓ Provides detailed explanatory context✓ Mentions testing or verification✓ Names security-relevant behavior explicitly
Why it was queued
signing boundarysigning or wallet path
AI analysis · Informational 14/100
This commit is a code cleanup and test reorganization. It moves sanity-check assertions (which verify that hardcoded cryptocurrency addresses and function signatures are correct) out of the main firmware code and into dedicated unit tests. The actual checks still run during development testing, but they no longer get baked into debug firmware builds. There is no security vulnerability being fixed here.
There is no need for calling `confirm_tx_data` and then `confirm_data_and_summary`, since we can do calldata streaming (while hashing and confirming) in `confirm_tx_data`.
[no changelog]
85/100 · StrongMessage clarity
✓ Specific, descriptive subject✓ Names a concrete action or component✓ Uses a recognizable type or scope✓ Provides detailed explanatory context
Why it was queued
signing or wallet pathsecond-pass: security-sensitive path
AI analysis · Low 31/100
This is a code cleanup (refactor) in Trezor's Ethereum transaction signing. It merges two internal steps into one so that transaction data is streamed, hashed, and shown to the user in a single pass instead of two. The commit message says there is no functional change, and the visible code preserves the same hashing and confirmation behavior. UI test snapshots changed, which is expected because the on-screen flow order changed, but no security bug is evident from the diff alone.
AI review queuedfeat(nordic): add support for different NCS sdk versionsby tychovrahe · 253c4f87 · Jul 16, 2026 · 10 filesMessage 62 · AdequateInformational 19Details
Commit message · tychovrahe
feat(nordic): add support for different NCS sdk versions
[no changelog]
62/100 · AdequateMessage clarity
✓ Specific, descriptive subject✓ Names a concrete action or component✓ Uses a recognizable type or scope! No meaningful explanatory body
Why it was queued
signing or wallet pathsecond-pass: security-sensitive path
AI analysis · Informational 19/100
This commit updates Trezor's Bluetooth firmware build system so it can compile against two different Nordic SDK versions (an older 2.9 and a newer 3.3). It adds version checks, pins the correct compiler toolchain, and renames a few Bluetooth and SPI constants so the same source code works with both SDKs. There is no obvious security bug being fixed; it is primarily a compatibility and build-reliability change.
- this commit adds clear signing support for Trezor Steakhouse ETH Prime Vault
72/100 · AdequateMessage clarity
✓ Descriptive subject✓ Names a concrete action or component✓ Uses a recognizable type or scope✓ Provides an explanatory body
Why it was queued
signing boundary
AI analysis · Informational 15/100
This commit adds metadata for a new Ethereum yield vault called 'Trezor Steakhouse ETH Prime Vault' so that Trezor devices can display a friendly name and token details when users interact with it. It is a product feature addition, not a security fix or vulnerability.
AI review queuedrefactor(core/ethereum): handle staking/yielding in `confirm_tx_data()`by Roman Zeyde · 6c158b50 · Jul 15, 2026 · 4 filesMessage 85 · StrongLow 27Details
Commit message · Roman Zeyde
refactor(core/ethereum): handle staking/yielding in `confirm_tx_data()`
This way, we don't need to call `get_progress_indicator()`, since all the data has been already loaded and hashed.
[no changelog]
85/100 · StrongMessage clarity
✓ Specific, descriptive subject✓ Names a concrete action or component✓ Uses a recognizable type or scope✓ Provides detailed explanatory context
Why it was queued
signing or wallet pathsecond-pass: security-sensitive path
AI analysis · Low 27/100
This is an internal code cleanup in Trezor's Ethereum transaction signing. It changes how staking and yield-farming transactions hand off their user-confirmation step so the confirmation happens inside the same routine instead of being returned as a separate progress indicator. The commit message says the goal is to avoid needing a progress indicator because all transaction data is already loaded and hashed. There is no direct evidence in the commit of a security bug being fixed, but the change touches the code that decides what users see before they approve high-value operations like staking and vault deposits/withdrawals. The large set of updated UI test hashes shows the on-screen flow changed, which is expected for a refactor of this kind.
✓ Descriptive subject✓ Names a concrete action or component✓ Uses a recognizable type or scope! No meaningful explanatory body
Why it was queued
cryptography-sensitive path
AI analysis · Low 44/100
This commit tightens validation in the hardware wallet's AES-GCM encryption code by rejecting initialization vectors (IVs) of length zero. Previously, a zero-length IV was silently accepted, which violates the NIST standard and can lead to weak or broken encryption. The change also makes the higher-level encrypt/decrypt functions properly report this error instead of ignoring it.
✓ Descriptive subject✓ Names a concrete action or component✓ Uses a recognizable type or scope✓ Mentions testing or verification! No meaningful explanatory body
Why it was queued
cryptography-sensitive path
AI analysis · Informational 12/100
This commit only adds a new test to the Trezor firmware crypto test suite. It checks that multiplying certain special 'low-order' points on the Curve25519 curve by various secret scalars always produces an all-zero result. There is no change to the actual cryptographic code being tested, only a new regression test. By itself, this commit does not fix or introduce a security vulnerability.
✓ Descriptive subject✓ Names a concrete action or component✓ Uses a recognizable type or scope! No meaningful explanatory body
Why it was queued
cryptography-sensitive path
AI analysis · Informational 15/100
This commit only updates the project's cryptographic test suite to match a newer version of the Wycheproof test-vector submodule. It changes test data file names and field names (for example, 'keyDer' to 'publicKeyDer') and adds a safety check for empty padding data in a test helper. None of these changes affect the actual Trezor firmware code that users rely on for security.
Security candidatefeat(crypto): test AES-GCM, AES-CCM, HMAC and PBKDF2 against wycheproof vectorsby Ondřej Vejpustek · 6bee661b · Jul 15, 2026 · 1 fileMessage 72 · AdequateInformational 12Details
Commit message · Ondřej Vejpustek
feat(crypto): test AES-GCM, AES-CCM, HMAC and PBKDF2 against wycheproof vectors
[no changelog]
72/100 · AdequateMessage clarity
✓ Specific, descriptive subject✓ Names a concrete action or component✓ Uses a recognizable type or scope✓ Mentions testing or verification! No meaningful explanatory body
Why it was queued
cryptography-sensitive path
AI analysis · Informational 12/100
This commit only adds and improves test code. It extends an existing test suite to check more cryptographic algorithms (AES-GCM, AES-CCM, HMAC, PBKDF2) using well-known public test vectors from Project Wycheproof. It also fixes a few small helper functions in the test file so they correctly report failures. There is no change to the actual Trezor firmware or its cryptographic library, so this commit does not introduce or fix a security vulnerability in the product itself.
Currently, the length of all bounded strings is checked using len(str). This returns the number of Unicode code points. However, the XDR spec defines string object<m> as a sequence of at most m bytes, see https://datatracker.ietf.org/doc/html/rfc4506#section-4.11. A single Unicode code point can consist of multiple bytes.
[no changelog]
88/100 · StrongMessage clarity
✓ Descriptive subject✓ Names a concrete action or component✓ Uses a recognizable type or scope✓ Provides detailed explanatory context✓ Links an issue, advisory, or supporting reference
Why it was queued
signing or wallet pathsecond-pass: security-sensitive path
AI analysis · Moderate 60/100
This commit fixes a length-check bug in Trezor's Stellar cryptocurrency support. The device was measuring string length in characters (Unicode code points) instead of bytes. Because some characters use multiple bytes, a string could pass the old check yet be too long for the Stellar protocol, potentially causing malformed transaction data or unexpected device behavior when signing.
Co-authored-by: PrisionMike <su.sh2396@gmail.com> Co-authored-by: Roman Zeyde <roman.zeyde@satoshilabs.com>
90/100 · StrongMessage clarity
✓ Specific, descriptive subject✓ Names a concrete action or component✓ Uses a recognizable type or scope✓ Provides detailed explanatory context✓ Names security-relevant behavior explicitly
Why it was queued
signing boundaryaccess controlboot or update path
AI analysis · Informational 15/100
This commit only adds new protocol buffer message definitions for an upcoming Ethereum feature (EIP-7702 authorization signing). It does not contain any actual signing logic, device firmware behavior, or user-interface code. It is a schema-only change and therefore does not introduce a security vulnerability by itself.
style(core/rust): fix clippy::cast_lossless where possible
[no changelog]
50/100 · ThinMessage clarity
✓ Specific, descriptive subject✓ Names a concrete action or component! No meaningful explanatory body
Why it was queued
cryptography-sensitive pathboot or update pathauthentication pathparser or protocol path
AI analysis · Informational 15/100
This is a code-style cleanup commit in the Trezor firmware's Rust code. It replaces plain `as` type casts with explicit conversion methods like `u32::from(...)`, `i16::from(...)`, and `f32::from(...)` to satisfy the Rust Clippy lint `cast_lossless`. These changes do not alter program behavior; they only make the code clearer and avoid compiler warnings. There is no security fix or functional change here.
AI review queuedrefactor(core/ethereum): rename and move keccak-related helper functionsby Roman Zeyde · dc2aadf6 · Jul 15, 2026 · 7 filesMessage 62 · AdequateInformational 15Details
Commit message · Roman Zeyde
refactor(core/ethereum): rename and move keccak-related helper functions
[no changelog]
62/100 · AdequateMessage clarity
✓ Specific, descriptive subject✓ Names a concrete action or component✓ Uses a recognizable type or scope! No meaningful explanatory body
Why it was queued
signing or wallet pathsecond-pass: security-sensitive path
AI analysis · Informational 15/100
This is a clean internal code cleanup in the Ethereum app of the Trezor firmware. It moves the Keccak hashing helper to a shared location and updates callers to use it. There is no user-facing behavior change and no security fix.
AI review queuedrefactor(core/ethereum): simplify `confirm_data_and_summary()`by Roman Zeyde · 814203e1 · Jul 14, 2026 · 2 filesMessage 77 · AdequateInformational 15Details
Note that `(confirm_data_chunk is None) == (confirm_summary is None)`.
[no changelog]
77/100 · AdequateMessage clarity
✓ Specific, descriptive subject✓ Names a concrete action or component✓ Uses a recognizable type or scope✓ Provides an explanatory body
Why it was queued
signing or wallet pathsecond-pass: security-sensitive path
AI analysis · Informational 15/100
This is a small internal code cleanup in Trezor's Ethereum transaction signing. It replaces two separate return values with a single tuple, because they are always either both present or both absent. There is no user-visible behavior change and no indication of a security fix.
format_duration_ms and format_autolock_duration implemented the same unit-selection logic but disagreed above 24 hours: the ApplySettings confirmation dialog formatted a 2-day delay as "48 hours" while the device menu showed "2 days" for the same setting. The day tier was added only to the device menu code (a57fc3f8ed) and never made it back to the older helper.
To reproduce (T3W1): set a day-scale delay from the host, e.g. `trezorctl set auto-lock-delay 48h` - the confirmation dialog reads "48 hours" - then check Settings -> Security -> Auto-lock in the device menu, which reads "2 days" for the same stored value.
Add the day tier to format_duration_ms, move the unit names inside it, and drop format_autolock_duration entirely, so all call sites share one implementation. Since the unit-name strings are not specific to the auto-lock feature, rename plurals__lock_after_x_minutes to just plurals__minutes etc., keeping their string IDs so existing translation blobs remain compatible.
Also add the day tier to the click-test oracle in tests/click_tests, which mirrored the day-less variant and would have mispredicted the device menu contents for day-scale delays.
[no changelog]
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
95/100 · StrongMessage clarity
✓ Descriptive subject✓ Names a concrete action or component✓ Uses a recognizable type or scope✓ Provides detailed explanatory context✓ Mentions testing or verification✓ Names security-relevant behavior explicitly
Why it was queued
boot or update path
AI analysis · Informational 21/100
This commit fixes a minor user-interface inconsistency in how the Trezor device displays auto-lock delays longer than a day. Previously, the confirmation dialog showed "48 hours" while the settings menu showed "2 days" for the same delay. The fix merges two similar formatting functions into one shared implementation so all screens use the same wording. It is a cosmetic bug fix, not a security vulnerability.
test(clear_signing): add Trezor Test descriptor and clear-signing tests
- Add test descriptors. - core tests.
[no changelog]
87/100 · StrongMessage clarity
✓ Specific, descriptive subject✓ Names a concrete action or component✓ Uses a recognizable type or scope✓ Provides an explanatory body✓ Mentions testing or verification
Why it was queued
signing boundarysigning or wallet path
AI analysis · Informational 15/100
This commit only adds two new import names (DateFormatter and RawFormatter) to an existing Python file in the Ethereum clear-signing module. It is a test-related change with no functional code modifications, no bug fixes, and no security-sensitive behavior changes.
- constant path (resolved by offline parser) - renamed `hint` to `is_mono` in clear signing [no changelog]
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
85/100 · StrongMessage clarity
✓ Specific, descriptive subject✓ Names a concrete action or component✓ Uses a recognizable type or scope✓ Provides detailed explanatory context
Why it was queued
signing boundarysigning or wallet path
AI analysis · Low 27/100
This commit adds a new way for Ethereum transaction definitions to include fixed text labels (called 'constant values') that are shown to the user during signing. It does not change how transaction data is parsed or signed, but it expands what untrusted display-format definitions can put on the device screen. The change is a feature addition, not a fix for a known bug, and there is no vendor statement that it is security-relevant.