Repository navigation
Releases
Releases · milesroxas/skill-studio
Skill Studio 0.17.0
Immutable
release. Only release title and notes can be modified.
Added
- Split up, among the three rewrites of Suggest changes, for a rule that asks for several things: one rule for each, each with its reason. It takes the place of Shorter on such a rule, where fewer words would not fix it. The status line says "Split rule 04 into 3 rules.", and each new rule takes its own number.
- The editor underlines a phrase that adds words and changes nothing, such as "in order to" or "make sure to", and several words where one says the same, such as "is able to" for "can".
Changed
- Code in a code block is colored: the language's own words, such as
constorreturn, are violet, values such as text in quotes and numbers are blue, and comments are grey. This works for JavaScript, TypeScript, CSS, HTML, JSON, Python, YAML and shell commands. - Changes, on the Skill screen, shows what changed since the last version, so it is no longer empty after a save. While the file has unsaved text, The last save beside it shows only that.
- After you save an edit, Edited since v1 stands over the versions on the Skill screen and under the skill's name. Click it to see what changed. Before, the app showed "v2 · not run yet", as if saving had made v2: a version is made by a run.
- The app says a run keeps a version, and the Assistant's button for one is Keep. Save is only for a file.
- A rule has no word limit. The editor no longer marks a rule for being over 40 words: a rule is as long as what it says needs. An amber number now always names something you can change, and it goes when you change it.
- Shorter makes a rule exact, not only brief: it cuts words that change nothing, says each thing once, and uses the one word where several stood for it.
- A vague word such as "reasonably" is now marked in a rule that also says "each" or "every", or "always": none of them says how much. A number still answers it.
- Shorter says "Nothing to cut" when every word in the rule does work, and says how many words it cut when it cuts. Before, it could offer a rewrite that was no shorter.
- Auto refine and the Best practice rewrite are checked before you see them, and written once more when they did not fix what the editor marked. So replacing a rule clears its amber number. When a rewrite cannot supply something only you know, such as the reason for a rule, it says what it needs and offers Help me improve.
- Auto refine no longer rewrites a rule whose trouble is not its wording: a line that changes nothing, a rule for another skill, or a long file. It says so and offers Help me improve.
- A red number reads the rule on the last version that ran it. Before, it counted every run there had been, so a rule stayed red after a later version's answers followed it.
- A draft in Help me improve says it meets the guide only when the editor's own checks find nothing in it either.
Fixed
- A table in a skill shows as a table in Formatted. Before, it showed its bars and dashes. Click a cell to edit it as text.
- A code block in Formatted no longer shows its ``` lines until you click in it. Its language, such as "css", stays as a small label over the code.
- A rule that says "otherwise", or that opens with its reason, is again marked when it holds too many conditions or asks for several things.
- A word a rule only names, in single quotes or as code, is no longer marked as a spare word. So a rule such as "Never write
please" keeps no amber number for it. - Changes says "No changes in this file since v1" when the open file is the same and another file was deleted or replaced. Before, it said nothing had changed.
- Edited since v1 no longer opens a file that the last version did not keep.
- Best practice no longer says "Still misses" for points Jev was not sure of, or for ones no rewrite can fix. The same goes for "Misses a point of the guide" under a draft in Help me improve.
- Suggest changes no longer warns "May drop a fact" on a rewrite that only adds a number or a reason to the rule.
- A long rule that asks for one thing is no longer marked as asking for several because of the commas in its reason.
- On a small window, the rewrites and Auto refine's revision no longer run off the top of the window.
- A rule whose reason follows a comma and "so", such as "..., so every shot matches", is no longer marked as having no reason.
- An underline in a rule now stands under the word the check read, not under the same letters inside a longer word.
Full Changelog: v0.16.1...v0.17.0
Skill Studio 0.16.1
Immutable
release. Only release title and notes can be modified.
Changed
- A note above the editor or a case, such as the one that offers Add quotes, puts its buttons at its right edge when there is room, so it no longer leaves an empty strip beside the words.
- Such a note marks what it needs from you with a colored dot before its title, as the rest of the app does. Its title is no longer colored, and it wraps rather than being cut off.
Full Changelog: v0.16.0...v0.16.1
Skill Studio 0.16.0
Immutable
release. Only release title and notes can be modified.
Added
- The Skill screen lists every file of the skill, not only SKILL.md and its references: scripts, other Markdown files and anything else the skill holds. Each text file can be edited. A picture or another file that is not text is listed and can be deleted.
- The files are listed by folder, and a click on a folder folds its files away. Add a file and Add a folder, beside Files, add one next to the open file, so a file can now sit beside SKILL.md. You type its name where it appears in the list.
- Right-click a file or a folder on the Skill screen to add a file or a folder there, or to delete the file.
- The editor says when a skill's name holds
anthropicorclaude, or its description holds a tag such as<example>: Anthropic refuses both in a skill uploaded to it.
Changed
- Restore puts back every file of a version, scripts and pictures included, so the files match the version again. Before, it put back SKILL.md and the references and left the rest.
- A version's page compares every file of the skill, and says when a picture or another file that is not text changed.
- The notes about a skill's name and description say what some apps will not take, such as a name that is not the folder's. Before, they read as if the skill could not be used.
- The Assistant can read and change a skill's scripts and other text files.
Fixed
- A file's changes no longer end with an empty line the file does not have.
- A skill whose description holds a colon, such as
description: Use when: a release ships, no longer shows as a file whose top lines cannot be read. The editor reads it as Claude Code does and offers Add quotes, so other apps can read it too. - A SKILL.md written by Write with AI, the selection tools or the assistant puts quotes around a description that holds a colon. A new skill's description stays on one line.
Full Changelog: v0.15.0...v0.16.0
Skill Studio 0.15.0
Immutable
release. Only release title and notes can be modified.
Added
- Recently deleted, under the versions on the Skill screen. It lists the versions and runs you deleted in the last 30 days, each with Undo, so you can bring one back after the notice is gone or the app restarts.
- A rule most answers break shows a red number in the editor, and pointing at any rule's number says how many answers broke it.
- The editor says when a rule asks for so many things that one answer cannot say which was missed.
- Click the number of cases under Lift to see how it is counted: which cases, the score without the skill, how many attempts each side is over, and how many rules Jev was not sure of. An opened case says how many attempts its own score is over.
Changed
- Delete version always asks first, even for a version with no runs. Undo on the status line still brings it back.
- A change in score now counts only the rules and checks marked on both runs, and says how many it left out. Before, a rule Jev was unsure of on one run and not the other could read as the skill getting better or worse.
- An opened case, an opened run and the lift's note say how many rules and checks Jev was not sure of behind a score. They stay out of the score until you mark them.
- A case's change, the lift and the change between two outputs count your own marks only on the rules and checks you marked on both runs, so a case no longer rises just because you checked its newest run more closely. Each score shown on its own still counts every mark you made.
- Jev now reads the text files a case gives the agent, such as a design system's notes, when it checks rules like "use only colours from the project".
- Grading again asks Jev only the questions whose answer could have changed: add a check and only that check is asked, change a check's weight or a case's title and nothing is. The first grading after this update asks every question once more.
- Reading the skill before writing is still shown and flagged, but no longer counts in Process, so a run with the skill and one without are scored on the same things.
- Lift is now the number, how far it moved and how many cases it is over, beside its ribbon. The sentences that stood beside the number are in the note the number of cases opens, which also says when the score without the skill was graded on another version's rules.
- The Run panel says when the Cursor agent will use every tool in a case that allows only some, and when an older version was kept without its scripts.
Fixed
- A skill kept in a repository no longer has that repository's
CLAUDE.mdor other skills read by every attempt. Each attempt works in a folder of its own, away from the skill, and its files are moved into the run when it ends. - An attempt that hit the time limit, could not start, or ended with an error (signed out, a usage limit) no longer counts as an answer. It has no score, it is left out of every average, and Results says why.
st runexits 1 when one fails. - A run stopped partway no longer scores a case from fewer attempts than it was asked for. The case has no score from that run until it is run again.
- A run cut off when the app quit or crashed says "not finished" under Runs.
- Any file a skill writes is now an output. A
.csv,.xmlor.ipynbis graded as text, and a.pdfor.docxshows no score instead of 0. - An answer that writes several files is graded on the one the case's prompt names, and Results says which file the score is for.
- A rule that does not come up for a case, such as "When the request names a table…" on a request that names none, shows under Does not apply instead of counting as met.
- A bullet under an example heading, or one that only says to read a file, is no longer graded as a rule. A skill written as numbered steps has its steps graded, and an answer with nothing to mark says why it has no score.
- Results says when Jev read only the start of a long answer, a file beside it or the prompt, so a rule broken past that point is no longer counted as followed. Jev also reads more of each than before.
- An answer graded by an older version of Skill Studio's questions, by another version of Jev, or without its page drawn has no score until you regrade its run, instead of standing beside scores counted another way. Scores graded before this update stand as they are.
- Allowing a tool or changing a case's time limit makes it a new request, and changing its turn or time limit asks for a regrade, so neither reads as the skill's doing.
- A check about a picture reads the picture as it was when the run was made, after the case is changed or deleted.
- A run whose record of its cases is damaged says so on Results, instead of quietly reading its answers against the cases as they are now.
- A version only one other model ran keeps its score, and its line names that model. One that several other models ran shows no score instead of an average that can't be compared.
- Running an older saved version gives the agent its scripts and other files.
- A run without the skill no longer takes a version number, so the files still count as changed until you run them with the skill.
- A case's hidden files, such as
.eslintrc.jsonor.github/, reach the agent. - An animated picture, a JPEG saved to load in passes, or a photo saved on its side no longer stops every rule from being graded. The app draws it again, or leaves it out and says why.
- When part of a grading fails, the answers that did come back are kept, and grading again asks only the rest.
- A grading that asks Jev nothing new keeps what the answers cost, and
st gradesays how much of it this grading paid. - Grading a run in the app while
stgrades it in a terminal is refused, with who is grading it, and no answer is lost to the other's write. st reportandst comparescore as Results does: your corrections count, a changed rules list or transcript setting asks for grading again, andst comparelists the answers it does not compare, with why.st self-testchecks a bad anchor against every good anchor of its case, and fails one that has no good anchor to lose to.- A case's page tells you when an answer was graded before rules and checks were marked on its kind of file, and that Regrade marks them.
- With an unsaved edit, the top bar's next step is to save it, not to run a version that doesn't have your edit.
- Restore warns you before it drops edits you haven't saved, and keeping those edits afterwards can no longer quietly undo the restore.
- Deleting a run or version no longer ends the Undo of the one you deleted before it, in this skill or another.
- Unsaved edits made in the app and in a browser on the same skill no longer overwrite each other.
- The editor says when the lines at the top of SKILL.md cannot be read, and when the name or description breaks a limit, instead of saying they are missing.
- Turning a skill off for an app works when its link has a different name from the skill's folder.
- The assistant no longer pays again for the whole skill each time you move to another screen.
- Help me improve says when TypeSafe refuses your key, and when Jev could not read an answer.
Full Changelog: v0.14.0...v0.15.0
Skill Studio 0.14.0
Immutable
release. Only release title and notes can be modified.
Added
- Keyboard shortcuts you can change. 1, 2 and 3 go to Skill, Cases and Results, R opens Run, and ? shows every shortcut. Change any of them in Settings, under Keyboard shortcuts: click one and press the new keys. A shortcut without Cmd (Ctrl on Windows) waits while you type in a field.
- Grade N runs on All cases, when some answers have no score they could have. One click grades every run that needs it. A case whose newest run only waits to be graded says Needs grading, and offers to grade that run when you click it.
Changed
- A long list of versions on the Skill screen shows the newest five. The older ones fold under one line you click to open.
- All cases on Results is read by case, not by run. Each case is one row: its last run, its score, and how the score changed from the run before. Before, every run was a column, so a skill with many runs scrolled sideways and a run of one case left Not run in every other row.
- You no longer have to run every case each time. A case is compared with its own earlier runs, whatever versions and models they used.
- Lift, at the top of All cases, sets the average of the cases' first scores against the average of their latest. Before, it compared two whole runs, and a run of one case stood for the whole skill.
- Under the lift, a line says how many cases have run on the version a run would test now. While some have not, it is a dark band with a button, such as Run the other 4 on v3, that opens Run with only those cases ticked. A case that has run on that version and only waits to be graded is not run again: the band says how many wait, and grades them once no other case is left.
- A score from before your last edit to the skill is drawn as an outline with no fill, on All cases and in the list on the left, until that case runs again.
- Cases are grouped: the ones checked on the current version, then the ones that are not. In each group the cases that dropped come first.
- Click a case on All cases to see where it stands in one sentence, the first things its answer missed, the next step, and its other runs.
- A long list of cases shows the first 8 of each group, with Show more and a Find a case field. The column heads stay in view as you scroll. Runs shows the newest 5, and each run says which case it answered.
- A case's change is read against its last run by the same model where there is one, so a switch of model is not read as a change in the skill.
- A run remembers the cases it answered. When you change a case's prompt or its files, the answers written before keep their scores and are shown apart, under Answered before the case changed on All cases. They are no longer compared with answers to the new prompt, and each shows the prompt it was given. The same holds for a case you delete.
- An answer has no score while its grading is out of date: after you add or change a check, after you change what the skill is graded as, or when it was graded without Jev and you have since added a key. It says which, and Grade brings the score back. Before, old and new scores were shown side by side as if they measured the same thing.
- A score over several answers shows only once every one of them is graded. Before, an answer whose grading had failed was left out without a word, so a score could cover fewer answers than it seemed to.
Removed
- The line of rules and checks added up for the latest version, at the top of All cases. A version's detail on the Skill screen still shows it.
Fixed
- Grading a run again no longer holds an old answer to a prompt you edited after it was written.
- Lift and a case's change are read only between answers to the same prompt, graded the same way: both by Jev, or both without.
- A version's score is read from one model across versions whenever that model ran two of them on a case in common. Before, an early run that shared no case with the latest stood in the way.
Full Changelog: v0.13.4...v0.14.0
Skill Studio 0.13.4
Immutable
release. Only release title and notes can be modified.
Changed
- Stopping a run no longer takes over Results. A case keeps showing its last finished answers and their scores, and a version whose run you stopped is not counted as the latest.
- A stopped answer is left out of the lists that pick which answers to compare. You can still read it by opening its run under Runs on All cases.
- After you stop a run before any answer is finished, the top bar no longer asks you to read its results.
Fixed
- A stopped answer is no longer shown in red as if it had failed. It says once, in plain words, that it was stopped before it finished.
Full Changelog: v0.13.3...v0.13.4
Skill Studio 0.13.3
Immutable
release. Only release title and notes can be modified.
Changed
- Cost to run on a case is now Run size, from small to very large. The bars are the same; the name no longer reads as a price, since what a run costs also turns on the model you pick. What the last run cost still shows beside it.
Full Changelog: v0.13.2...v0.13.3
Skill Studio 0.13.2
Immutable
release. Only release title and notes can be modified.
Added
- Cost to run on a case: five bars, from light to very heavy, that move as you write it. Pictures, files and an answer that is a page or a picture add the most; checks and a good answer add almost nothing. Point at the bars to see what adds to the cost and what an attempt cost in the last run.
- In a wide window, Cases shows the last answer to a case beside its form, with its score and what its run cost, so checks can be written with the answer in view. Wider still, the form splits in two: the task, then a good answer, the checks and the length.
Changed
- A case's optional parts fold under two lines, A good answer and checks and Length and size. A closed line says what it holds, such as "7 checks".
Full Changelog: v0.13.1...v0.13.2
Skill Studio 0.13.1
Immutable
release. Only release title and notes can be modified.
Fixed
- Each answer in a run keeps its working files in its own folder,
scratch, beside its outputs. Before, they went to a temporary folder that every run shared, so one answer could pick up files another had left.
Full Changelog: v0.13.0...v0.13.1
Skill Studio 0.13.0
Immutable
release. Only release title and notes can be modified.
Added
- File to grade on a case, for an answer that comes as both a page and a picture: pick the one you would use, and only that file is graded. It shows once a run of the case has written both, at the top of the case's good answer, checks and size.
- A picture made from a page is measured through that page. When the picture is the file graded and the page beside it draws the same picture, a rule or check about a size, a distance or a position is marked from the page's real numbers. A picture the page does not draw is still judged by eye.
Changed
- With File to grade set, the size check on a case says which file it is about: "Check the picture’s size in pixels" or "Set the page’s size in pixels".
Full Changelog: v0.12.0...v0.13.0