saffron-ai 1.3.0 → 1.3.1

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
package/CHANGELOG.md CHANGED
@@ -5,6 +5,533 @@ the VS Code extension (`ChathurangaJayasinghe.saffron-vscode`) and the
5
5
  JetBrains plugin (`ai.saffron.jetbrains`). Dates are publish dates. Each
6
6
  release heading names its main feature: `## <version>: <title>`.
7
7
 
8
+ ## 1.3.1: Checks for what vanishes, URL and style checks, and a StepSet accepted on its own
9
+
10
+ Released 2026-11-02.
11
+
12
+ Runner
13
+
14
+ Upgrading from 1.3.0:
15
+
16
+ - A recording with a negative or exact URL check, a check of a message
17
+ that vanishes, a computed-style check, or a value captured earlier
18
+ inside the element a step looks for is cache schema version 14, 15, 16
19
+ or 17, which 1.3.0 and earlier refuse ("Update saffron-ai"). An Outline
20
+ recording whose cell is a `{data:...}` value holding a date is version
21
+ 18. Upgrade every machine and CI job that replays the suite together.
22
+ - `saffron.config.json` refuses a quoted number (`"actionTimeoutMs":
23
+ "10000"`), a negative one, or a `screenshot` other than `off`,
24
+ `viewport` or `full-page`, where it used to take them as they were.
25
+ - A row, data or captured value inside an API request's path is
26
+ percent-encoded where it would change the URL (`#`, `?`, a space; `&`,
27
+ `=` and `+` in the query): `A#1` is sent as `A%231`, no longer cut at
28
+ the `#`. A `/` and an existing `%20` are sent as they are.
29
+ - A `*` that opens a scenario after a Background ending in a `Then` is an
30
+ action now, as Cucumber reads it, not a check: a recording of such a
31
+ step records once more.
32
+ - A data table or doc string written under a `StepSet <name>` call is an
33
+ error naming the file and line, where it used to be dropped.
34
+ - With `reset` commands, recordings run one at a time when `agentWorkers`
35
+ is above 1 and the agent is on, and the run says so: one session's
36
+ reset would wipe the data another is recording on. Replays still use
37
+ `workers` as before.
38
+ - `strict`, `verifyProposals` and `reuseSteps` in `saffron.config.json`
39
+ take `true` or `false` without quotes; a quoted `"false"` used to read
40
+ as on.
41
+ - In an Outline, a number or boolean a request body or JSON check holds
42
+ only by coincidence with a row's value (a field not named like the
43
+ column, or a step not naming it) is kept as recorded and binds the
44
+ recording to its row, where it used to be sent and checked as each
45
+ row's value. Other rows are refused with a message until the field is
46
+ named like the column or the step names it. A constant beside a column
47
+ the step writes (`active: true` beside `admin: <admin>`) is kept as
48
+ recorded without binding the row.
49
+ - The recorder refuses checks that cannot fail: a URL check every URL
50
+ holds (`/`), a text, style, download or response pattern that matches
51
+ any text (`.*`), and a watch for a message other than the one the
52
+ `Then` names.
53
+ - A watched message counts as seen only when it appears after the watch
54
+ was armed (an element already showing it must stop showing it once),
55
+ and a seen check uses its watch up: a second check of the same message
56
+ needs a watch of its own. A recording that relied on a message left on
57
+ screen goes red and heals.
58
+ - `saffron diff --json` prints a list for one proposal too: a script that
59
+ read an object for one proposal reads the first entry now.
60
+ - SIGTERM stops a run as Ctrl-C does and exits 143, not 2.
61
+ - `--workers 0`, `--agent-workers 0` and a worker count that is not a
62
+ whole number exit 2 instead of running one worker.
63
+ - `init --examples` no longer writes into an existing
64
+ `saffron.config.json`.
65
+
66
+ - **From real-world use of Saffron.**
67
+ - A data value an API request takes through its step's table (`|
68
+ lastName | {data:signup.lastName} |`, sent as `<table:lastName>`) can
69
+ change and the recording replays at zero AI: the freshness check
70
+ follows the table cell to the data it names, and a header name HTTP
71
+ defines (`User-Agent`) is never taken for an old value ("User").
72
+ (SAF-177)
73
+ - A URL check with no URL is refused while recording, saying that a URL
74
+ check reads what the URL contains, and that there is no "does not
75
+ contain" or "is exactly" URL check yet: four such checks had been
76
+ recorded empty, and only the paid proof refused them. (SAF-178)
77
+ - `saffron run --plan --json` prints only the JSON on stdout, with a
78
+ shard too: the `shard 1/2 ...` line goes to stderr. (SAF-179)
79
+ - Fewer false sensitivity notes: a check is judged against the last step
80
+ that acted on the page, a step that only waits passed over; a
81
+ StepSet's exit guard is not probed against the set's own last action,
82
+ as it says the whole set got where it should (a set of checks only is
83
+ probed against the step before it); and a step that reads a whole
84
+ body, page or message gets no `broad check:` note. (SAF-180)
85
+ - From a review of these fixes: `--plan --json` also keeps the note on
86
+ Examples rows that share a recording off stdout, and an empty shard's
87
+ plan is an empty JSON plan, with no shard report written; a StepSet
88
+ used twice in a row has each use's exit guard left unprobed; an old
89
+ data value written into a body key or a custom header's name, whole or
90
+ in part, still makes the recording stale; "the body of a card", "the
91
+ entire row" and a card's body are not read as the whole page; and a URL
92
+ check's text is recorded trimmed. A header's value is still searched
93
+ for an old data value (`Authorization: Bearer <old token>`); only the
94
+ header names HTTP defines are left out, and a header an extract reads
95
+ by name is not taken for data.
96
+
97
+ - **From a three-round review of the runner.**
98
+ - No false green from a blank row value: a role name pattern, a response
99
+ URL, a JSON field or header pattern, a watched message and a style
100
+ pattern that a blank value leaves matching more than they mean are
101
+ refused, as text and URL checks already were. An expected status left
102
+ blank is refused, never read as any 2xx.
103
+ - Outline rows never record over each other's proposal: a row whose
104
+ pending proposal was recorded for another row is red, with a committed
105
+ recording too, and a proposal from another machine never turns a
106
+ failing protected `Then` yellow.
107
+ - A fixture takes values only from a proof that passed, and a proposal
108
+ from another machine is never replayed for its values.
109
+ - A scenario whose run throws (a busy proposal lock, a full disk) is that
110
+ scenario's red; the run goes on and writes its report. Reports, shard
111
+ reports and the history are written whole, never half.
112
+ - A contract check repaired by a heal keeps every operand of the check:
113
+ a URL, pattern or attribute changed is review, not repair.
114
+ - Secrets: a value set with a trailing newline is masked as the page
115
+ shows it, trimmed; credential headers and cookies in a trace are
116
+ replaced whole (`Authorization: Basic ...`); names written as one word
117
+ (`authtoken`, `JSESSIONID`) count as credentials; a header the
118
+ transcript keeps is judged by the same rule as the trace.
119
+ - A pattern holding a harmless escape (`\#`, `\:`) next to a row value
120
+ no longer fails as an invalid regular expression.
121
+ - `{captured:...}` in the element a step looks for is the captured value,
122
+ and a retried "message was shown" check can still see the message.
123
+ - API scenarios: YAML merge keys (`<<: *base`) in a contract are applied,
124
+ the response schema is chosen by the response's content type,
125
+ `differs` compares lists and objects, `the /x field is null` compiles,
126
+ a phrase-compiled `Then` serves every Outline row, and with
127
+ `api.transcript: "off"` no error quotes a response body.
128
+ - Recording: a multi-line typed text keeps its `{captured:...}`, a
129
+ `{unique:...}` inside an Examples cell is typed resolved, a feature
130
+ edit writes `<plan>` only for the whole word (`Production` stays),
131
+ and Antigravity stops when the run aborts.
132
+ - `saffron eject` writes tests never looser than replay: an exact path
133
+ or style is compared whole, enabled and disabled checks need the
134
+ element shown, `differs` compares amounts, a not-visible check reads
135
+ every locator at once, timed waits are capped, the device's own
136
+ engine is dropped, and Playwright's trace is left out when the tests
137
+ read `{env:...}` values, since it would record them unmasked.
138
+ - CLI: `run --plan` validates its flags as `run` does, `coverage --write
139
+ --json` prints only JSON, the HTML report escapes every proposal
140
+ field, and `init` sets up the rest before naming an MCP file it
141
+ cannot merge.
142
+ - Amounts and dates: `(1.5 kg)` is positive, and a date range that ends
143
+ before it starts is refused.
144
+
145
+ - **From a second two-round review of the runner.**
146
+ - Outline rows: `accept --propagate` no longer copies a healed locator
147
+ holding its Outline row's value into other recordings, and `saffron
148
+ status` marks an Outline stale when any row would record again, naming
149
+ the row.
150
+ - Secrets: a value set with a trailing newline is masked where the page
151
+ holds it trimmed (in a URL, in another case, escaped in a selector or
152
+ glued into a word), and a warning about an unhandled rejection masks
153
+ the credentials a session captured.
154
+ - Recording: a target with nothing to find the element by is refused, a
155
+ `saffron_step` whose text names another step is refused so actions are
156
+ never filed a step late, the first `saffron_done` verdict stands, a
157
+ pass past the time budget that leaves a step unrecorded is a failure,
158
+ and a Claude session stopped at its turn limit says to raise
159
+ `maxTurns`; Claude token counts include compaction.
160
+ - Files and processes: `prune` keeps a merge-conflicted recording of a
161
+ live scenario outside `features` and the recording of an Outline whose
162
+ Examples have no rows for now; a scenario name of any length can be
163
+ saved, and recordings saved under the whole name are still found;
164
+ Ctrl-C stops a running `reset`; a `hookTimeoutMs` past a timer's limit
165
+ no longer ends every hook at once; a hook timed out on Windows ends with
166
+ its child processes; a run killed while writing history no longer
167
+ spoils the next line; a browser that fails to close or a history file
168
+ that cannot be written no longer costs the run its report; writes ride
169
+ out a file briefly locked on Windows; a failed browser close leaves no
170
+ video folder behind; `init` git-ignores the temp and lock files a
171
+ killed run leaves.
172
+ - `saffron eject` writes tests never looser than replay: a schema check
173
+ finds its operation with the row's value in the path, a response with
174
+ a schema per JSON media type is refused, a blank expected status and
175
+ the NAT64 and IPv4-compatible link-local forms are refused, attribute
176
+ and style checks read the element replay reads, a blank captured value
177
+ in a checked text fails, a dialog answer from a row is that row's, a
178
+ whitespace-only text pattern is fixme, and header patterns compile as
179
+ replay compiles them.
180
+ - Status, plan and coverage: a feature in another Gherkin language shows
181
+ its name, `run --plan` counts a feature's fixture, `coverage` refuses a
182
+ suite with a file it cannot parse and `verify` says so, and a scenario
183
+ named like an Outline row keeps its own last result.
184
+ - Editors and reports: the language server reads doc string lines as
185
+ data, not StepSet calls, and the CI run panel shows a cost under a cent
186
+ as the report does.
187
+
188
+ - **From a two-round review of the runner.**
189
+ - No false green from another row or a leftover: an Outline cell that
190
+ reads a date through a data value has its day checked as a date cell
191
+ does; a step reused by swapping a quoted value no longer offers the old
192
+ value's other recordings; a message already on screen before an action
193
+ no longer counts as seen; a capture pattern a blank row value leaves
194
+ open is refused.
195
+ - Recording types a captured value as plain text, as replay does, and
196
+ refuses a browser call whose captured value holds an `{env:...}`
197
+ token rather than type the secret.
198
+ - Checks without a frame also find what an iframe attached after the
199
+ check began shows; checks with fallback locators end at their timeout;
200
+ a drag finds its drop zone by its fallbacks; the heal guard runs a
201
+ check on the tab an earlier action moved to or opened.
202
+ - Seeded recording keeps a watch armed in one stretch for its `Then` in
203
+ the next, and seeded steps and the sensitivity check read values
204
+ another step's table or doc string gives.
205
+ - API: a schema check finds the operation for a path with a parameter
206
+ inside a segment (`/files/{name}.json`); a row value after a captured
207
+ URL that brings its own query is encoded as a query value, in replay
208
+ and in exported tests; with no agent, an Outline's rows file one
209
+ contract re-check, not one each; `api pin` passes over a fallback URL
210
+ that does not answer with OpenAPI 3.
211
+ - Proposals: `accept` notices a hand edit of a proposal's API contract
212
+ fingerprints, `accept --step-set` refuses to replace another StepSet
213
+ whose name makes the same file name, and `run` reports two StepSets or
214
+ a fixture and a scenario whose recordings would share a file; an
215
+ accepted heal no longer copies a captured value into other scenarios;
216
+ a "capture differs" note leaves credentials out.
217
+ - Reports: a password holding an `@` in a URL on the command line is
218
+ hidden whole, `report --merge` refuses a shard it cannot show, naming
219
+ the field, an exported or imported bundle's report names only the
220
+ evidence the bundle carries, and the trace page marks a request the
221
+ agent did not record.
222
+ - CLI and editors: `saffron author` fails with the provider's error
223
+ instead of writing it as the draft; `init` refuses an instructions
224
+ block with no end marker instead of deleting the text after it; a
225
+ config file holding something other than an object is refused by
226
+ name; an exported timed wait written with a row value waits as long
227
+ as replay; feature edits apply to steps whose keyword is several
228
+ words (`Étant donné que`); the language server started by an editor
229
+ reads a project with no `features` key as the default folder.
230
+
231
+ - **A `Then` is read-only while recording.** A check refuses a browser
232
+ action that enters, presses keys, picks, ticks, uploads, drags or
233
+ navigates, and a click on a button that submits or commits (Log in,
234
+ Save, Submit, Confirm and the like), before the browser runs it: the
235
+ agent is told the step is a check and to fail it, suggesting the action
236
+ as a `When`, if it does not hold as the page stands. Before, such a
237
+ click was recorded and proved, and held for review afterwards; a login
238
+ suite had a check click Log in, and others submit again to bring back a
239
+ toast that had vanished. A click that opens what is checked (a menu, a
240
+ tab) is still recorded and held as before, and proposals filed before
241
+ this stay held from `accept --all`. (SAF-181)
242
+
243
+ - **A "not shown" check must see something disappear.** The sensitivity
244
+ check after a proof runs each "not shown" check in a `Then` that holds
245
+ other checks on its own, before the step it follows; one that holds
246
+ there too (its element was never on the page, as with an invented
247
+ selector) gets an `insensitive check:` note naming it, held from
248
+ `accept --all`. Before, the step's other checks failing there hid it,
249
+ and it stayed green whatever the page showed. The recording agent is
250
+ also told to target the element a named field labels (the booker, not
251
+ any paragraph showing the name) and only elements it saw for a "not
252
+ shown" check. (SAF-183)
253
+
254
+ - **A StepSet's own data value stays its own in an Outline.** A step that
255
+ enters a `{data:...}` value itself, while not naming the column (a
256
+ StepSet typing `{data:guests.first.surname}`), no longer has the rows
257
+ refused because one row's cell reads the same: the email row of a login
258
+ suite was refused for it. A value a step only checks is still refused
259
+ for the rows that differ, as the page may show it from the row.
260
+ (SAF-182)
261
+
262
+ - **Trial navigations stay out of the recording.** A navigation the
263
+ agent leaves for the same page with another query or fragment, before
264
+ doing anything there, is discarded and noted in the proposal: a login
265
+ suite kept each query format the agent tried as a replay action. A
266
+ navigation to another page, or one acted on, is kept. (SAF-191)
267
+
268
+ - **A part of a name the agent wrote replays.** A click through a locator
269
+ the agent wrote, naming a part of the element's name, is recorded with
270
+ the full name the page it saw last shows for the one element of that
271
+ role holding it: replay matches a name whole, and timed out on "Sign
272
+ in". Where the page showed none or several, the proposal notes it
273
+ while recording, before the proof. (SAF-184)
274
+
275
+ - **Negative and exact URL checks.** A URL check can now say that the
276
+ URL does not contain a value, or that its path is exactly one, recorded
277
+ from the step's own words: before, a URL check was containment only,
278
+ and steps saying the URL no longer holds the login could not be
279
+ recorded. Replay and `saffron eject` check them alike. A recording
280
+ holding one is cache schema version 14, so an older runner refuses it
281
+ instead of checking containment. (SAF-186)
282
+
283
+ - **Checks for messages that vanish.** A toast or error flash gone
284
+ before its `Then` runs can be checked: the agent arms a watch
285
+ (`saffron_watch`) in the `When` step before the action that shows it,
286
+ and the `Then` checks that it was seen since. Replay watches the same
287
+ way. Before, a login error toast vanished before the check, and
288
+ recordings submitted again inside the `Then` to bring it back. A
289
+ recording holding a watch is cache schema version 15; `saffron eject`
290
+ writes its scenario as `test.fixme` for now. (SAF-185)
291
+
292
+ - **Computed-style checks.** A step can check an element's computed CSS
293
+ property (an invalid input's `border-color` is `rgb(220, 38, 38)`), or
294
+ match it to a pattern: before, only classes, attributes and the visible
295
+ error could be checked. Replay and `saffron eject` check it alike, in
296
+ Chromium, Firefox and WebKit. A recording holding one is cache schema
297
+ version 16. `saffron eject` also writes the negative and exact URL
298
+ checks now, which it had written as `test.fixme`. (SAF-187)
299
+
300
+ - **A navigation to a captured URL stays the capture.** When the agent
301
+ goes to `{captured:signin}` and the value is an address on the base URL,
302
+ the recording keeps `{captured:signin}`: before, it kept the day's path
303
+ relative to the base, and replay went to the old link instead of the one
304
+ its own request got. (SAF-188)
305
+
306
+ - **Accept one StepSet from a proposal.** `saffron accept <file>
307
+ --step-set "Sign in"` files only that set's steps, which its proof
308
+ replayed green, for every other scenario to reuse at zero AI, while the
309
+ proposal and its failing check stay pending. Before, a failed check kept
310
+ a verified sign-in from being reused, and each scenario paid to record it
311
+ again. A proposal whose proof failed records how many steps passed before
312
+ it (`meta.provedSteps`), and a set the proof did not get through is
313
+ refused. (SAF-189)
314
+
315
+ - **Reuse picks the recording the page's state fits.** A step recorded in
316
+ two page states (a promo field opened first, or found open) keeps both
317
+ recordings, and seeding replays the one whose controls the page shows
318
+ now: before, it always took the newest, which closed a field already open
319
+ and paid the agent to record the step again. (SAF-190)
320
+
321
+ - **CLI and files, from the 1.0 review.** (SAF-133)
322
+ - A scenario with no steps is red, saying so, and starts no agent
323
+ session: it had recorded nothing and passed.
324
+ - A broken link or a folder named `*.json` under `.saffron/proposals`
325
+ no longer stops `status`, `diff` and the end of a run.
326
+ - `saffron diff --json` prints a list for one proposal too, as it does
327
+ for several.
328
+ - `accept`, `reject` and `diff` find a project-relative file (as
329
+ `status` prints it) under `--project-root`, not only from the current
330
+ folder.
331
+ - `accept --with-feature-edit` edits a step line that ends in spaces.
332
+ - `saffron status` prints an unreadable recording's warning on a line of
333
+ its own.
334
+ - `saffron init` keeps a skill you edited under `.saffron/backups/skills/`
335
+ before installing the shipped one.
336
+ - Already fixed and checked: an empty run exits 2 naming why, a StepSet
337
+ file that does not parse is named, recordings are written atomically,
338
+ a CRLF file takes feature edits, and status counts an unreadable
339
+ recording apart.
340
+
341
+ - **A check's selector is tried on the page while recording.** A check
342
+ whose CSS selector matches nothing on the page the agent sees is refused
343
+ then, at no cost, naming the likely cause (a link styled as a button is
344
+ an `<a>`): before, only the paid proof caught `p:has(button)`. The other
345
+ half of this story, a `broad check:` note for a check held by a whole
346
+ card, shipped in 1.3.0. (SAF-171)
347
+
348
+ - **Exact `Then` phrases need no agent.** In an API scenario, `Then the
349
+ response should answer 404`, `Then the /status field is active` and
350
+ `Then the response matches the Customer schema` are recorded as those
351
+ checks without the agent, which records only the steps around them. Any
352
+ other wording goes to the agent as before, never guessed. An option, not
353
+ a required syntax. (SAF-162)
354
+
355
+ - **Live progress, and a time budget per agent session.** While an agent
356
+ records, the run prints on stderr each step it records and each proof as
357
+ it ends (`· Checkout: recording step 41 of 92: ...`, `· Checkout: proof
358
+ 1 passed`): a 92-step recording ran 340 seconds with no word until it
359
+ finished. `maxSessionMinutes` (`--max-session-minutes`) ends one session
360
+ that has gone on that long: its tools refuse, and it ends unfinished
361
+ with the agent's account of where it got to. (SAF-194)
362
+
363
+ - **Row proofs show their time and each row's latest result.** `saffron
364
+ diff` and the review page say when an Outline's rows were proved, and
365
+ beside each row its latest result in the last run (`latest run: red at
366
+ ...`, or `no run since`): a proposal's passing row proofs were read as
367
+ readiness while later runs of those rows timed out. (SAF-192)
368
+
369
+ - **A failed response is named beside the step it broke.** A step that
370
+ fails after one of the app's own requests reported failure (an error
371
+ status, or a JSON body with `success: false`, `ok: false` or an
372
+ `error`) says so in its error, which the console, the report and the
373
+ heal prompt all show: an HTTP 200 holding "overbooked" before a timeout
374
+ reaching checkout no longer reads like a locator problem. (SAF-193)
375
+
376
+ - **Outline rows, from the 1.0 review.** (SAF-128)
377
+ - A list check whose values come from `{data:...}` or `{env:...}` cells
378
+ ("Alice, Bobby" from `{data:guests.a}` and `{data:guests.b}`) is tied
379
+ to its row, as one of plain cells is, so another row fails it with the
380
+ reason and no agent session: it was written as the row's `<column>`,
381
+ and every other row failed it and paid the agent again on every run.
382
+ - The guide says what a row-bound recording keeps of a short secret
383
+ read through an Examples cell, and why a PIN belongs elsewhere.
384
+ - Checked and kept with tests: a value inside a word of what a
385
+ Background step types (`test` in `qa-tester@example.com`) binds no row;
386
+ a date cell typed as ISO, a value seen only as a slug or in another
387
+ case, and a column naming a selector's role refuse the other rows
388
+ instead of recording again; a URL-encoded value is the row's
389
+ `<column>`; a `{data:...}` value naming an unset variable stops the run
390
+ before it starts; a short secret glued to other text is masked.
391
+
392
+ - **Visible text, from the 1.0 review.** (SAF-129)
393
+ - A check whose value is written on several lines (a two-line doc
394
+ string) holds where the page draws those lines, split by a `<br>` or
395
+ by blocks: it never matched "Line one<br>Line two", which read "Line
396
+ oneLine two". A value on one line still reads the text as one run, so
397
+ two blocks drawn apart never make it.
398
+ - A list check (`expectEach`) no longer passes on a hidden option, or
399
+ hidden text, that differs from a shown one only in case: the page's
400
+ lines in another case count only where CSS `text-transform` changes
401
+ them.
402
+ - A check on a large container is faster: 22 ms instead of 56 ms on a
403
+ 4,000-row section. 0.9.9 takes 4 ms there, as it did not read what is
404
+ drawn; the rest is the cost of reading only drawn text.
405
+ - Kept as it is: an element without text whose first match is hidden
406
+ gives way to a displayed one, as a hidden mobile copy does to the
407
+ desktop one; a check that means one row's badge targets that row.
408
+
409
+ - **CLI, reports and the trace viewer, from the 1.0 review.** (SAF-130)
410
+ - `saffron diff` names the text a role with no name is found by:
411
+ `expect listitem with text "Milk" to be visible`, not every listitem.
412
+ - The trace viewer titles another browser session's step by the keyword
413
+ it was written with (`Given`, `And`), not by its kind, and a session
414
+ opens on a step of its own rather than on another session's first one.
415
+ - The results page embeds at most 20 MB of screenshots, failures first;
416
+ past that it names each file instead (300 failures made a 78 MB page).
417
+ - A feature reached through a linked folder's alias, or named in another
418
+ case on a case-insensitive disk, is the suite's feature, with its
419
+ recordings, not a new one.
420
+ - The MCP tools keep answering while `saffron.config.json` cannot be read,
421
+ on the config they last read, and say what is wrong with it.
422
+ - `init --examples` leaves an existing `saffron.config.json` as it is:
423
+ the examples' `retries`, `strict` and `trace` no longer change the
424
+ project's own runs, and a missing `baseURL` is named in a note.
425
+ - A bundle made in a git checkout with no remote is refused for that, not
426
+ as one made outside a checkout.
427
+ - SIGTERM stops a run as Ctrl-C does (teardown, the report of what
428
+ finished) and exits 143, instead of a red line and exit 2.
429
+ - A red scenario that replayed no step is labelled `not run` in the run
430
+ output, not `cache`.
431
+ - `--workers 0`, or a typo in `--agent-workers` or in the config's
432
+ `workers` or `agentWorkers`, exits 2 instead of running one worker.
433
+ - Fewer temporary folders are left behind by the test suite.
434
+ - Kept as they are: `saffron status` already marks an Examples row a
435
+ row-bound recording refuses (✗, with the reason), and the chronic hint
436
+ counts only reds that cost an agent session, which a refused row never
437
+ does. `saffron report --no-open` already writes the report without
438
+ opening a browser (1.3.0).
439
+
440
+ - **Language server and editor parity, from the 1.0 review.** (SAF-131)
441
+ - The language server reads data Examples under every Examples keyword
442
+ of the file's language (`Scenarios:`, `Exemples:`), as the parser
443
+ does, instead of flagging them; a bare `{data:file}` in a step is no
444
+ reference to the runner, so it gets no hover and no error; and the
445
+ `# language:` header is read with Gherkin's own pattern.
446
+ - Its completion rates a step after `As "<who>"` (a session an Examples
447
+ column names) unrecorded, as step reuse never seeds it.
448
+
449
+ - A text check that fails on a value the page holds but does not show says
450
+ where it sits: `Timed out waiting for text "Total: 42 NOK": the page holds
451
+ it, but not shown (inside a closed <details>)`, or in an element with
452
+ `display: none`, the `hidden` attribute, `visibility: hidden` or
453
+ `content-visibility: hidden`, naming the element. A total that is gone and one behind a panel that did not open read apart. Text
454
+ the page does not hold reads as before. (SAF-195)
455
+ - A text check on a large container is fast again: 3.5 ms on a
456
+ 4,000-row container (0.9.9 took 2.4 ms, reading hidden text too). The value is looked for in the container's text first,
457
+ and only the text it covers is asked whether it is drawn; anything that
458
+ question cannot settle (a textarea, a dropdown, a shadow host) is read
459
+ the full way, so the check passes on exactly what it passed on before.
460
+ (SAF-129)
461
+ - A range of days the page shows with its year (`20–22 Oct 2026`,
462
+ `Oct 30 - Nov 1, 2026`, `31 Dec 2026–2 Jan 2027`, German
463
+ `20.-22.10.2026`), up to 62 days long, is recorded as a range token
464
+ (`{date+7–+9:D MMM YYYY}`), as a single day already was. A step that
465
+ names a stay by intent ("check-in in 7 days for 2 nights") now records
466
+ a check that follows the clock instead of a date frozen on the day of
467
+ recording. The agent is no longer told to match dates with a pattern,
468
+ which passes on any day. (SAF-124)
469
+ - The agent skill and the block `saffron init` writes into `AGENTS.md` and
470
+ `CLAUDE.md` say when a small supporting script is the right tool (only
471
+ for a value Saffron cannot express, asked for first) and how to keep it
472
+ trustworthy: expected values never taken from the output being checked,
473
+ run as a `setup` command writing git-ignored `data/` files, assumptions
474
+ stated, and a report of what could replace it. Run `saffron init` again
475
+ to update the block. (SAF-125)
476
+ - A video of each scenario's run: `"video": "retain-on-failure"` keeps
477
+ the video of every scenario that is not green, `"on"` every one, in
478
+ `video.webm` beside its screenshots; also `--video` for one run. Off
479
+ by default. It works in Chromium, Firefox and WebKit, and a heal or a
480
+ recording is in the video of the run it happened in; proof replays are
481
+ not filmed. A browser that cannot record (no ffmpeg) stops the run once,
482
+ at the start. A summary export carries no video. A video shows what is
483
+ typed, secrets included, so keep it off for scenarios that type them.
484
+ (SAF-51)
485
+ - The limits of the CI runs viewed into `.saffron/ci/` are settings:
486
+ `"ci": {"cacheRuns": 20, "cacheMB": 4096}`. The defaults stay 10 runs
487
+ and 1 GB. (SAF-121)
488
+ - Each feature edit a proposal suggests names the file and line it
489
+ rewrites, in `saffron diff` (`file` and `line` in `--json`), the review
490
+ page and the run report. A StepSet's step names the StepSet's
491
+ definition, where accept rewrites it for every scenario using the set,
492
+ and an edit to a Background step is marked skipped, as accept leaves it
493
+ to you. (SAF-31)
494
+ - `saffron status --json` lists each pending proposal's feature edits
495
+ (`featureEdits`, as `diff --json` gives them, with file and line), and
496
+ the VS Code and JetBrains panels show them in a proposal's details
497
+ without another command. (SAF-30)
498
+ - Each step a proposal changes is marked with its class: `repair` (only
499
+ how the step finds the same element again or waits for it: a fallback
500
+ selector, a timeout, a wait), `review` (what it does changed, an
501
+ element named another way, or one of Saffron's notes names it) or, for a
502
+ suggested feature edit, `protected` (the wording, which only you
503
+ apply). It shows in `saffron diff` (`adaptation` in `--json`, with
504
+ counts in `summary`), the review page and the run report, and the
505
+ guide has a section on the three classes. (SAF-32)
506
+ - From a review of this release: a step whose target is named by a
507
+ selector inside an iframe is `review` when the selector changes (the
508
+ frame names the iframe, not the element); the heal branch's pull request
509
+ body, the run report and the editors' trace panel class each step as
510
+ `saffron diff` does, reading the proposal's notes, and name the file and
511
+ line of each feature edit; a video with no viewport setting is filmed at
512
+ the page's own 1280x720, not Playwright's 800x800 fit; a range that
513
+ writes the month twice (`Oct 20 - Oct 22, 2026`), one ending exactly 31
514
+ days back, or one from before today to after it (`Room 7 - 22 Oct 2026`
515
+ is a room) is kept whole, never with only its end rewritten.
516
+
517
+ Editors (VS Code extension 0.7.7, JetBrains plugin 0.7.8)
518
+
519
+ - Both editors list the feature file edits a proposal suggests, each with
520
+ the file and line `saffron accept --with-feature-edit` rewrites and why
521
+ accept would skip one: VS Code in the proposal tooltip, JetBrains in the
522
+ Proposals tab details, read from `saffron status --json`. (SAF-30)
523
+ - VS Code: a `# language:` header Gherkin rejects is a plain comment; a
524
+ file a link inside the test folder leads back to is listed once; a
525
+ StepSet under a nested `saffron.config.json` in the test folder is
526
+ defined. (SAF-131)
527
+ - JetBrains: the Files tab fails on a folder it cannot read, as the runner
528
+ does, instead of hiding it; right-click Run reads the IDE's view of the
529
+ files, not the disk through links, each time the menu opens; a Windows
530
+ junction is walked as a link. (SAF-131)
531
+ - Both editors open a report path through a link outside the test folder
532
+ only when it really leads into the project, as the runner follows links
533
+ inside the test folder only. (SAF-131)
534
+
8
535
  ## 1.3.0: Recordings checked by machine, and a review every view agrees on
9
536
 
10
537
  Released 2026-10-30.