Skip to content

DON'T SQUASH: Improve skills - #470

Open
kisvegabor wants to merge 3 commits into
masterfrom
skills2
Open

kisvegabor wants to merge 3 commits into
masterfrom
skills2

Conversation

@kisvegabor

Copy link
Copy Markdown
Member

No description provided.

@cubic-dev-ai cubic-dev-ai Bot left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

15 issues found across 106 files

Prompt for AI agents (unresolved issues)

Check if these issues are valid — if so, understand the root cause of each and fix them. When an issue isn't valid or won't be fixed in this PR, reply in its thread with the reason and then resolve the thread. If appropriate, use sub-agents to investigate and fix each issue separately.


<file name="examples/lvgl_pro/.agents/skills/idea-to-xml/SKILL.md">

<violation number="1" location="examples/lvgl_pro/.agents/skills/idea-to-xml/SKILL.md:47">
P2: This LICENSE.md requirement has no step of its own, and it sits before the approved spec, so the agent may create it too early or never. It also splits the sources lead-in from its table. Move it into step 4 as a bullet, as `image-to-xml/SKILL.md` does.</violation>
</file>

<file name="examples/lvgl_open/.agents/skills/write-readme/SKILL.md">

<violation number="1" location="examples/lvgl_open/.agents/skills/write-readme/SKILL.md:214">
P3: This rule conflicts with the Targets item below it. Target names are `#define` constants (`<PROJECT>_TARGET_<NAME>`), but item 7 asks for their values. State which target names belong in the README, or exempt them from this rule.</violation>
</file>

<file name="examples/lvgl_open/.agents/skills/idea-to-xml/SKILL.md">

<violation number="1" location="examples/lvgl_open/.agents/skills/idea-to-xml/SKILL.md:47">
P3: This adds a root `LICENSE.md` while the paragraph above still asks for a license file next to each asset and a source in the spec. Say which record is required, or whether the new file replaces the others, so the agent does not keep three copies.</violation>
</file>

<file name="templates/basic/.agents/skills/write-readme/SKILL.md">

<violation number="1" location="templates/basic/.agents/skills/write-readme/SKILL.md:116">
P2: The emsdk location and the clone behavior are not supported by the repo docs. `docs/cli.mdx` documents the CLI cache as `~/.config/lvgl-editor/resources`. Remove the path and clone claim, and keep only the statement that the docs' CI workflow has no emsdk step.</violation>

<violation number="2" location="templates/basic/.agents/skills/write-readme/SKILL.md:222">
P2: Step 5 assumes generated files exist, but step 1 allows writing the README without them. Make step 5 conditional on step 1 having run the CLI, so the XML fallback names are not deleted.</violation>
</file>

<file name="templates/common/.agents/skills/image-to-xml/SKILL.md">

<violation number="1" location="templates/common/.agents/skills/image-to-xml/SKILL.md:68">
P3: The `LICENSE.md` bullet asks for a source and license for every asset but does not say what to write when they are unknown. Add a fallback such as writing `unconfirmed` and listing the asset in the report, so the Report item has a defined marker to point to.</violation>
</file>

<file name="templates/basic/.agents/skills/image-to-xml/SKILL.md">

<violation number="1" location="templates/basic/.agents/skills/image-to-xml/SKILL.md:70">
P2: Step 4 requires a license for every asset, but step 7 expects some licenses to stay unconfirmed. Say what to write when the license is unknown, for example `unknown`, so the agent does not guess.</violation>
</file>

<file name="templates/common/.agents/skills/idea-to-xml/SKILL.md">

<violation number="1" location="templates/common/.agents/skills/idea-to-xml/SKILL.md:47">
P2: The new `LICENSE.md` instruction conflicts with the existing rule to put a license file next to each asset. Say which one is required, or whether both are. A root `LICENSE.md` is also usually the project's own license, so this name may collide with it.</violation>
</file>

<file name="templates/basic/.agents/skills/idea-to-xml/SKILL.md">

<violation number="1" location="templates/basic/.agents/skills/idea-to-xml/SKILL.md:47">
P2: This LICENSE.md instruction sits in the Ask step, before the spec is approved, so it asks the agent to create a file before "Build nothing before the user approves it" allows. The project's assets are not final at this point. Move it into step 4 next to the `globals.xml` bullets, as `image-to-xml` does.</violation>
</file>

<file name="tutorials/.agents/skills/idea-to-xml/SKILL.md">

<violation number="1" location="tutorials/.agents/skills/idea-to-xml/SKILL.md:56">
P2: The Photos hint says to scale with `<convert ... color_format="rgb565"/>`, but `color_format` does not resize the image. Show the size attribute too, for example `width="..."`, as `AGENTS.md` does.</violation>
</file>

<file name="templates/empty/.agents/skills/image-to-xml/recipes.md">

<violation number="1" location="templates/empty/.agents/skills/image-to-xml/recipes.md:151">
P2: This command crops from the original `ref.png`, but the text says to cut at display scale. Use the display-size copy `/tmp/img/ref.png` so the coordinates and output size match the displayed logo.</violation>
</file>

<file name="templates/common/.agents/skills/write-readme/SKILL.md">

<violation number="1" location="templates/common/.agents/skills/write-readme/SKILL.md:56">
P2: This tells the developer to change the `COMPILE_TARGET` `#define` in `<project>.h`, but that define is generated in `<project>_gen.h`, which `generate` overwrites. Say to define `<PROJECT>_COMPILE_TARGET` in `<project>.h` before its `#include` of `<project>_gen.h`.</violation>

<violation number="2" location="templates/common/.agents/skills/write-readme/SKILL.md:106">
P3: The token requirement is stated too broadly: `--help` and `--version` can still be used without a license token. Limit this to license-protected project commands so the skill can still check CLI availability and version.</violation>
</file>

<file name="examples/lvgl_pro/AGENTS.md">

<violation number="1" location="examples/lvgl_pro/AGENTS.md:524">
P2: The example for `trigger="{true}"` reads the event (`lv_event_get_current_target(e)`), but the section says that a `{ }` trigger has no real LVGL event, so the callback must not read it. Reconcile the two: either document that `e` is valid for `{true}` callbacks, or pass the view another way and drop the event read from this example.</violation>
</file>

<file name="tutorials/.agents/skills/write-readme/SKILL.md">

<violation number="1" location="tutorials/.agents/skills/write-readme/SKILL.md:104">
P2: These CLI behaviours are not in the repo docs, and the README this skill generates will repeat them as facts. Either cite a source for the token message, the exit code, and the emsdk clone, or mark them as unverified. The emsdk claim also conflicts with the editor docs, which say the user must install the emsdk tools.</violation>
</file>

Reply with feedback, questions, or to request a fix.

View guided diff | Turn on auto-fix | Re-trigger cubic

license of each asset, put the license file next to it, and write the source in the
spec.

Create a `LICENSE.md` file at the project root. List every image and font that the

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

P2: This LICENSE.md requirement has no step of its own, and it sits before the approved spec, so the agent may create it too early or never. It also splits the sources lead-in from its table. Move it into step 4 as a bullet, as image-to-xml/SKILL.md does.

Prompt for AI agents
Check if this issue is valid — if so, understand the root cause and fix it. When an issue isn't valid or won't be fixed in this PR, reply in its thread with the reason and then resolve the thread. At examples/lvgl_pro/.agents/skills/idea-to-xml/SKILL.md, line 47:

<comment>This LICENSE.md requirement has no step of its own, and it sits before the approved spec, so the agent may create it too early or never. It also splits the sources lead-in from its table. Move it into step 4 as a bullet, as `image-to-xml/SKILL.md` does.</comment>

<file context>
@@ -44,6 +44,10 @@ Then propose sources from this list. Ask before you download anything. Check the
 license of each asset, put the license file next to it, and write the source in the
 spec.
 
+Create a `LICENSE.md` file at the project root. List every image and font that the
+project uses, also the ones that were already there. For each one, give the file name,
+the source, and the license.
</file context>


### 5. Check the names

Search for each C name of the README in the generated files, for example

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

P2: Step 5 assumes generated files exist, but step 1 allows writing the README without them. Make step 5 conditional on step 1 having run the CLI, so the XML fallback names are not deleted.

Prompt for AI agents
Check if this issue is valid — if so, understand the root cause and fix it. When an issue isn't valid or won't be fixed in this PR, reply in its thread with the reason and then resolve the thread. At templates/basic/.agents/skills/write-readme/SKILL.md, line 222:

<comment>Step 5 assumes generated files exist, but step 1 allows writing the README without them. Make step 5 conditional on step 1 having run the CLI, so the XML fallback names are not deleted.</comment>

<file context>
@@ -0,0 +1,234 @@
+
+### 5. Check the names
+
+Search for each C name of the README in the generated files, for example
+`grep -rn "subject_volume" --include=*.h .`. Remove or fix every name that is not there.
+Then read the README from top to bottom once, as the firmware developer.
</file context>
Suggested change
Search for each C name of the README in the generated files, for example
Search for each C name of the README in the generated files (skip this when step 1 fell back to the XML; those names stay marked as not verified), for example

Comment on lines +116 to +117
- The CLI uses its own emscripten copy in `~/.config/lvgl-editor/emsdk`, and clones it
when it is missing. So CI needs no emsdk step.

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

P2: The emsdk location and the clone behavior are not supported by the repo docs. docs/cli.mdx documents the CLI cache as ~/.config/lvgl-editor/resources. Remove the path and clone claim, and keep only the statement that the docs' CI workflow has no emsdk step.

Prompt for AI agents
Check if this issue is valid — if so, understand the root cause and fix it. When an issue isn't valid or won't be fixed in this PR, reply in its thread with the reason and then resolve the thread. At templates/basic/.agents/skills/write-readme/SKILL.md, line 116:

<comment>The emsdk location and the clone behavior are not supported by the repo docs. `docs/cli.mdx` documents the CLI cache as `~/.config/lvgl-editor/resources`. Remove the path and clone claim, and keep only the statement that the docs' CI workflow has no emsdk step.</comment>

<file context>
@@ -0,0 +1,234 @@
+  reference PNGs, or CI creates new ones on every run and checks nothing. A failed
+  compare saves the screen next to the reference, with an `_err` suffix.
+- The tests run on the first target in `project.xml` only.
+- The CLI uses its own emscripten copy in `~/.config/lvgl-editor/emsdk`, and clones it
+  when it is missing. So CI needs no emsdk step.
+- In CI, store the token as an encrypted secret, for example a GitHub Actions secret
</file context>
Suggested change
- The CLI uses its own emscripten copy in `~/.config/lvgl-editor/emsdk`, and clones it
when it is missing. So CI needs no emsdk step.
- The CI workflow in `docs/cli.mdx` runs the CLI without an emsdk step, so do not add one.

before `<fonts>` or `<images>`. Keep the names that are already there.
- `LICENSE.md` at the project root: list every image and font that the project uses,
also the ones that were already there. For each one, give the file name, the source,
and the license.

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

P2: Step 4 requires a license for every asset, but step 7 expects some licenses to stay unconfirmed. Say what to write when the license is unknown, for example unknown, so the agent does not guess.

Prompt for AI agents
Check if this issue is valid — if so, understand the root cause and fix it. When an issue isn't valid or won't be fixed in this PR, reply in its thread with the reason and then resolve the thread. At templates/basic/.agents/skills/image-to-xml/SKILL.md, line 70:

<comment>Step 4 requires a license for every asset, but step 7 expects some licenses to stay unconfirmed. Say what to write when the license is unknown, for example `unknown`, so the agent does not guess.</comment>

<file context>
@@ -65,6 +65,9 @@ Later sessions read this section first. Update it when the UI changes.
   before `<fonts>` or `<images>`. Keep the names that are already there.
+- `LICENSE.md` at the project root: list every image and font that the project uses,
+  also the ones that were already there. For each one, give the file name, the source,
+  and the license.
 
 ### 5. Build the components, then the screens
</file context>
Suggested change
and the license.
and the license, or `unknown` if you cannot confirm it.

license of each asset, put the license file next to it, and write the source in the
spec.

Create a `LICENSE.md` file at the project root. List every image and font that the

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

P2: The new LICENSE.md instruction conflicts with the existing rule to put a license file next to each asset. Say which one is required, or whether both are. A root LICENSE.md is also usually the project's own license, so this name may collide with it.

Prompt for AI agents
Check if this issue is valid — if so, understand the root cause and fix it. When an issue isn't valid or won't be fixed in this PR, reply in its thread with the reason and then resolve the thread. At templates/common/.agents/skills/idea-to-xml/SKILL.md, line 47:

<comment>The new `LICENSE.md` instruction conflicts with the existing rule to put a license file next to each asset. Say which one is required, or whether both are. A root `LICENSE.md` is also usually the project's own license, so this name may collide with it.</comment>

<file context>
@@ -44,6 +44,10 @@ Then propose sources from this list. Ask before you download anything. Check the
 license of each asset, put the license file next to it, and write the source in the
 spec.
 
+Create a `LICENSE.md` file at the project root. List every image and font that the
+project uses, also the ones that were already there. For each one, give the file name,
+the source, and the license.
</file context>


Facts to state about the tests:

- The CLI runs them: `npm install --global @lvgl/lvglpro` (Node 18 or newer). It

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

P2: These CLI behaviours are not in the repo docs, and the README this skill generates will repeat them as facts. Either cite a source for the token message, the exit code, and the emsdk clone, or mark them as unverified. The emsdk claim also conflicts with the editor docs, which say the user must install the emsdk tools.

Prompt for AI agents
Check if this issue is valid — if so, understand the root cause and fix it. When an issue isn't valid or won't be fixed in this PR, reply in its thread with the reason and then resolve the thread. At tutorials/.agents/skills/write-readme/SKILL.md, line 104:

<comment>These CLI behaviours are not in the repo docs, and the README this skill generates will repeat them as facts. Either cite a source for the token message, the exit code, and the emsdk clone, or mark them as unverified. The emsdk claim also conflicts with the editor docs, which say the user must install the emsdk tools.</comment>

<file context>
@@ -0,0 +1,234 @@
+
+Facts to state about the tests:
+
+- The CLI runs them: `npm install --global @lvgl/lvglpro` (Node 18 or newer). It
+  needs an LVGL Pro license token in `LVGLPRO_CLI_TOKEN`, see the
+  [CLI docs](https://lvgl.io/docs/pro/cli). Without the token every
</file context>

your code in `<project>.c` or in files listed in `user_config.cmake`. The simulator
commands from `AGENTS.md`.

Do not list the component `_create` functions or the `#define` constants. A developer

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

P3: This rule conflicts with the Targets item below it. Target names are #define constants (<PROJECT>_TARGET_<NAME>), but item 7 asks for their values. State which target names belong in the README, or exempt them from this rule.

Prompt for AI agents
Check if this issue is valid — if so, understand the root cause and fix it. When an issue isn't valid or won't be fixed in this PR, reply in its thread with the reason and then resolve the thread. At examples/lvgl_open/.agents/skills/write-readme/SKILL.md, line 214:

<comment>This rule conflicts with the Targets item below it. Target names are `#define` constants (`<PROJECT>_TARGET_<NAME>`), but item 7 asks for their values. State which target names belong in the README, or exempt them from this rule.</comment>

<file context>
@@ -0,0 +1,234 @@
+    your code in `<project>.c` or in files listed in `user_config.cmake`. The simulator
+    commands from `AGENTS.md`.
+
+Do not list the component `_create` functions or the `#define` constants. A developer
+finds them in the headers.
+
</file context>

license of each asset, put the license file next to it, and write the source in the
spec.

Create a `LICENSE.md` file at the project root. List every image and font that the

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

P3: This adds a root LICENSE.md while the paragraph above still asks for a license file next to each asset and a source in the spec. Say which record is required, or whether the new file replaces the others, so the agent does not keep three copies.

Prompt for AI agents
Check if this issue is valid — if so, understand the root cause and fix it. When an issue isn't valid or won't be fixed in this PR, reply in its thread with the reason and then resolve the thread. At examples/lvgl_open/.agents/skills/idea-to-xml/SKILL.md, line 47:

<comment>This adds a root `LICENSE.md` while the paragraph above still asks for a license file next to each asset and a source in the spec. Say which record is required, or whether the new file replaces the others, so the agent does not keep three copies.</comment>

<file context>
@@ -44,6 +44,10 @@ Then propose sources from this list. Ask before you download anything. Check the
 license of each asset, put the license file next to it, and write the source in the
 spec.
 
+Create a `LICENSE.md` file at the project root. List every image and font that the
+project uses, also the ones that were already there. For each one, give the file name,
+the source, and the license.
</file context>

Comment on lines +68 to +70
- `LICENSE.md` at the project root: list every image and font that the project uses,
also the ones that were already there. For each one, give the file name, the source,
and the license.

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

P3: The LICENSE.md bullet asks for a source and license for every asset but does not say what to write when they are unknown. Add a fallback such as writing unconfirmed and listing the asset in the report, so the Report item has a defined marker to point to.

Prompt for AI agents
Check if this issue is valid — if so, understand the root cause and fix it. When an issue isn't valid or won't be fixed in this PR, reply in its thread with the reason and then resolve the thread. At templates/common/.agents/skills/image-to-xml/SKILL.md, line 68:

<comment>The `LICENSE.md` bullet asks for a source and license for every asset but does not say what to write when they are unknown. Add a fallback such as writing `unconfirmed` and listing the asset in the report, so the Report item has a defined marker to point to.</comment>

<file context>
@@ -65,6 +65,9 @@ Later sessions read this section first. Update it when the UI changes.
 - `globals.xml`: the tokens as `<consts>`, the `<fonts>`, the `<subjects>`, the
   `<images>`, and the shared `<styles>`, in this order: `validate` rejects `<styles>`
   before `<fonts>` or `<images>`. Keep the names that are already there.
+- `LICENSE.md` at the project root: list every image and font that the project uses,
+  also the ones that were already there. For each one, give the file name, the source,
+  and the license.
</file context>
Suggested change
- `LICENSE.md` at the project root: list every image and font that the project uses,
also the ones that were already there. For each one, give the file name, the source,
and the license.
- `LICENSE.md` at the project root: list every image and font that the project uses,
also the ones that were already there. For each one, give the file name, the source,
and the license. If you cannot confirm the source or the license, write `unconfirmed`
and list the asset in the report.

Comment on lines +106 to +107
[CLI docs](https://lvgl.io/docs/pro/cli). Without the token every
command stops with "No token provided" and exit code 1.

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

P3: The token requirement is stated too broadly: --help and --version can still be used without a license token. Limit this to license-protected project commands so the skill can still check CLI availability and version.

Prompt for AI agents
Check if this issue is valid — if so, understand the root cause and fix it. When an issue isn't valid or won't be fixed in this PR, reply in its thread with the reason and then resolve the thread. At templates/common/.agents/skills/write-readme/SKILL.md, line 106:

<comment>The token requirement is stated too broadly: `--help` and `--version` can still be used without a license token. Limit this to license-protected project commands so the skill can still check CLI availability and version.</comment>

<file context>
@@ -0,0 +1,234 @@
+
+- The CLI runs them: `npm install --global @lvgl/lvglpro` (Node 18 or newer). It
+  needs an LVGL Pro license token in `LVGLPRO_CLI_TOKEN`, see the
+  [CLI docs](https://lvgl.io/docs/pro/cli). Without the token every
+  command stops with "No token provided" and exit code 1.
+- Run `lvglpro generate .` first: the tests build the exported C, and a template does
</file context>
Suggested change
[CLI docs](https://lvgl.io/docs/pro/cli). Without the token every
command stops with "No token provided" and exit code 1.
[CLI docs](https://lvgl.io/docs/pro/cli). Without the token, a license-protected project command
stops with "No token provided" and exit code 1.

This branch has not been deployed

No deployments
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant