@mthanh/vespercli 1.7.0 → 1.8.0

package/assets/instructions/flutter/dart-n-flutter.instructions.md

@@ -0,0 +1,447 @@
+---
+description: 'Instructions for writing Dart and Flutter code following the official recommendations.'
+applyTo: '**/*.dart'
+---
+
+# Dart and Flutter
+
+Best practices recommended by the Dart and Flutter teams. These instructions were taken from [Effective Dart](https://dart.dev/effective-dart) and [Architecture Recommendations](https://docs.flutter.dev/app-architecture/recommendations).
+
+## Effective Dart
+
+Over the past several years, we've written a ton of Dart code and learned a lot about what works well and what doesn't. We're sharing this with you so you can write consistent, robust, fast code too. There are two overarching themes:
+
+1.  **Be consistent.** When it comes to things like formatting, and casing, arguments about which is better are subjective and impossible to resolve. What we do know is that being *consistent* is objectively helpful.
+
+    If two pieces of code look different it should be because they *are* different in some meaningful way. When a bit of code stands out and catches your eye, it should do so for a useful reason.
+
+2.  **Be brief.** Dart was designed to be familiar, so it inherits many of the same statements and expressions as C, Java, JavaScript and other languages. But we created Dart because there is a lot of room to improve on what those languages offer. We added a bunch of features, from string interpolation to initializing formals, to help you express your intent more simply and easily.
+
+    If there are multiple ways to say something, you should generally pick the most concise one. This is not to say you should `code golf` yourself into cramming a whole program into a single line. The goal is code that is *economical*, not *dense*.
+
+### The topics
+
+We split the guidelines into a few separate topics for easy digestion:
+
+*   **Style** – This defines the rules for laying out and organizing code, or at least the parts that `dart format` doesn't handle for you. The style topic also specifies how identifiers are formatted: `camelCase`, `using_underscores`, etc.
+
+*   **Documentation** – This tells you everything you need to know about what goes inside comments. Both doc comments and regular, run-of-the-mill code comments.
+
+*   **Usage** – This teaches you how to make the best use of language features to implement behavior. If it's in a statement or expression, it's covered here.
+
+*   **Design** – This is the softest topic, but the one with the widest scope. It covers what we've learned about designing consistent, usable APIs for libraries. If it's in a type signature or declaration, this goes over it.
+
+### How to read the topics
+
+Each topic is broken into a few sections. Sections contain a list of guidelines. Each guideline starts with one of these words:
+
+*   **DO** guidelines describe practices that should always be followed. There will almost never be a valid reason to stray from them.
+
+*   **DON'T** guidelines are the converse: things that are almost never a good idea. Hopefully, we don't have as many of these as other languages do because we have less historical baggage.
+
+*   **PREFER** guidelines are practices that you *should* follow. However, there may be circumstances where it makes sense to do otherwise. Just make sure you understand the full implications of ignoring the guideline when you do.
+
+*   **AVOID** guidelines are the dual to "prefer": stuff you shouldn't do but where there may be good reasons to on rare occasions.
+
+*   **CONSIDER** guidelines are practices that you might or might not want to follow, depending on circumstances, precedents, and your own preference.
+
+Some guidelines describe an **exception** where the rule does *not* apply. When listed, the exceptions may not be exhaustive—you might still need to use your judgement on other cases.
+
+This sounds like the police are going to beat down your door if you don't have your laces tied correctly. Things aren't that bad. Most of the guidelines here are common sense and we're all reasonable people. The goal, as always, is nice, readable and maintainable code.
+
+### Rules
+
+#### Style
+
+##### Identifiers
+
+*   DO name types using `UpperCamelCase`.
+*   DO name extensions using `UpperCamelCase`.
+*   DO name packages, directories, and source files using `lowercase_with_underscores`.
+*   DO name import prefixes using `lowercase_with_underscores`.
+*   DO name other identifiers using `lowerCamelCase`.
+*   PREFER using `lowerCamelCase` for constant names.
+*   DO capitalize acronyms and abbreviations longer than two letters like words.
+*   PREFER using wildcards for unused callback parameters.
+*   DON'T use a leading underscore for identifiers that aren't private.
+*   DON'T use prefix letters.
+*   DON'T explicitly name libraries.
+
+##### Ordering
+
+*   DO place `dart:` imports before other imports.
+*   DO place `package:` imports before relative imports.
+*   DO specify exports in a separate section after all imports.
+*   DO sort sections alphabetically.
+
+##### Formatting
+
+*   DO format your code using `dart format`.
+*   CONSIDER changing your code to make it more formatter-friendly.
+*   PREFER lines 80 characters or fewer.
+*   DO use curly braces for all flow control statements.
+
+#### Documentation
+
+##### Comments
+
+*   DO format comments like sentences.
+*   DON'T use block comments for documentation.
+
+##### Doc comments
+
+*   DO use `///` doc comments to document members and types.
+*   PREFER writing doc comments for public APIs.
+*   CONSIDER writing a library-level doc comment.
+*   CONSIDER writing doc comments for private APIs.
+*   DO start doc comments with a single-sentence summary.
+*   DO separate the first sentence of a doc comment into its own paragraph.
+*   AVOID redundancy with the surrounding context.
+*   PREFER starting comments of a function or method with third-person verbs if its main purpose is a side effect.
+*   PREFER starting a non-boolean variable or property comment with a noun phrase.
+*   PREFER starting a boolean variable or property comment with "Whether" followed by a noun or gerund phrase.
+*   PREFER a noun phrase or non-imperative verb phrase for a function or method if returning a value is its primary purpose.
+*   DON'T write documentation for both the getter and setter of a property.
+*   PREFER starting library or type comments with noun phrases.
+*   CONSIDER including code samples in doc comments.
+*   DO use square brackets in doc comments to refer to in-scope identifiers.
+*   DO use prose to explain parameters, return values, and exceptions.
+*   DO put doc comments before metadata annotations.
+
+##### Markdown
+
+*   AVOID using markdown excessively.
+*   AVOID using HTML for formatting.
+*   PREFER backtick fences for code blocks.
+
+##### Writing
+
+*   PREFER brevity.
+*   AVOID abbreviations and acronyms unless they are obvious.
+*   PREFER using "this" instead of "the" to refer to a member's instance.
+
+#### Usage
+
+##### Libraries
+
+*   DO use strings in `part of` directives.
+*   DON'T import libraries that are inside the `src` directory of another package.
+*   DON'T allow an import path to reach into or out of `lib`.
+*   PREFER relative import paths.
+
+##### Null
+
+*   DON'T explicitly initialize variables to `null`.
+*   DON'T use an explicit default value of `null`.
+*   DON'T use `true` or `false` in equality operations.
+*   AVOID `late` variables if you need to check whether they are initialized.
+*   CONSIDER type promotion or null-check patterns for using nullable types.
+
+##### Strings
+
+*   DO use adjacent strings to concatenate string literals.
+*   PREFER using interpolation to compose strings and values.
+*   AVOID using curly braces in interpolation when not needed.
+
+##### Collections
+
+*   DO use collection literals when possible.
+*   DON'T use `.length` to see if a collection is empty.
+*   AVOID using `Iterable.forEach()` with a function literal.
+*   DON'T use `List.from()` unless you intend to change the type of the result.
+*   DO use `whereType()` to filter a collection by type.
+*   DON'T use `cast()` when a nearby operation will do.
+*   AVOID using `cast()`.
+
+##### Functions
+
+*   DO use a function declaration to bind a function to a name.
+*   DON'T create a lambda when a tear-off will do.
+
+##### Variables
+
+*   DO follow a consistent rule for `var` and `final` on local variables.
+*   AVOID storing what you can calculate.
+
+##### Members
+
+*   DON'T wrap a field in a getter and setter unnecessarily.
+*   PREFER using a `final` field to make a read-only property.
+*   CONSIDER using `=>` for simple members.
+*   DON'T use `this.` except to redirect to a named constructor or to avoid shadowing.
+*   DO initialize fields at their declaration when possible.
+
+##### Constructors
+
+*   DO use initializing formals when possible.
+*   DON'T use `late` when a constructor initializer list will do.
+*   DO use `;` instead of `{}` for empty constructor bodies.
+*   DON'T use `new`.
+*   DON'T use `const` redundantly.
+
+##### Error handling
+
+*   AVOID catches without `on` clauses.
+*   DON'T discard errors from catches without `on` clauses.
+*   DO throw objects that implement `Error` only for programmatic errors.
+*   DON'T explicitly catch `Error` or types that implement it.
+*   DO use `rethrow` to rethrow a caught exception.
+
+##### Asynchrony
+
+*   PREFER async/await over using raw futures.
+*   DON'T use `async` when it has no useful effect.
+*   CONSIDER using higher-order methods to transform a stream.
+*   AVOID using Completer directly.
+*   DO test for `Future<T>` when disambiguating a `FutureOr<T>` whose type argument could be `Object`.
+
+#### Design
+
+##### Names
+
+*   DO use terms consistently.
+*   AVOID abbreviations.
+*   PREFER putting the most descriptive noun last.
+*   CONSIDER making the code read like a sentence.
+*   PREFER a noun phrase for a non-boolean property or variable.
+*   PREFER a non-imperative verb phrase for a boolean property or variable.
+*   CONSIDER omitting the verb for a named boolean parameter.
+*   PREFER the "positive" name for a boolean property or variable.
+*   PREFER an imperative verb phrase for a function or method whose main purpose is a side effect.
+*   PREFER a noun phrase or non-imperative verb phrase for a function or method if returning a value is its primary purpose.
+*   CONSIDER an imperative verb phrase for a function or method if you want to draw attention to the work it performs.
+*   AVOID starting a method name with `get`.
+*   PREFER naming a method `to...()` if it copies the object's state to a new object.
+*   PREFER naming a method `as...()` if it returns a different representation backed by the original object.
+*   AVOID describing the parameters in the function's or method's name.
+*   DO follow existing mnemonic conventions when naming type parameters.
+
+##### Libraries
+
+*   PREFER making declarations private.
+*   CONSIDER declaring multiple classes in the same library.
+
+##### Classes and mixins
+
+*   AVOID defining a one-member abstract class when a simple function will do.
+*   AVOID defining a class that contains only static members.
+*   AVOID extending a class that isn't intended to be subclassed.
+*   DO use class modifiers to control if your class can be extended.
+*   AVOID implementing a class that isn't intended to be an interface.
+*   DO use class modifiers to control if your class can be an interface.
+*   PREFER defining a pure `mixin` or pure `class` to a `mixin class`.
+
+##### Constructors
+
+*   CONSIDER making your constructor `const` if the class supports it.
+
+##### Members
+
+*   PREFER making fields and top-level variables `final`.
+*   DO use getters for operations that conceptually access properties.
+*   DO use setters for operations that conceptually change properties.
+*   DON'T define a setter without a corresponding getter.
+*   AVOID using runtime type tests to fake overloading.
+*   AVOID public `late final` fields without initializers.
+*   AVOID returning nullable `Future`, `Stream`, and collection types.
+*   AVOID returning `this` from methods just to enable a fluent interface.
+
+##### Types
+
+*   DO type annotate variables without initializers.
+*   DO type annotate fields and top-level variables if the type isn't obvious.
+*   DON'T redundantly type annotate initialized local variables.
+*   DO annotate return types on function declarations.
+*   DO annotate parameter types on function declarations.
+*   DON'T annotate inferred parameter types on function expressions.
+*   DON'T type annotate initializing formals.
+*   DO write type arguments on generic invocations that aren't inferred.
+*   DON'T write type arguments on generic invocations that are inferred.
+*   AVOID writing incomplete generic types.
+*   DO annotate with `dynamic` instead of letting inference fail.
+*   PREFER signatures in function type annotations.
+*   DON'T specify a return type for a setter.
+*   DON'T use the legacy typedef syntax.
+*   PREFER inline function types over typedefs.
+*   PREFER using function type syntax for parameters.
+*   AVOID using `dynamic` unless you want to disable static checking.
+*   DO use `Future<void>` as the return type of asynchronous members that do not produce values.
+*   AVOID using `FutureOr<T>` as a return type.
+
+##### Parameters
+
+*   AVOID positional boolean parameters.
+*   AVOID optional positional parameters if the user may want to omit earlier parameters.
+*   AVOID mandatory parameters that accept a special "no argument" value.
+*   DO use inclusive start and exclusive end parameters to accept a range.
+
+##### Equality
+
+*   DO override `hashCode` if you override `==`.
+*   DO make your `==` operator obey the mathematical rules of equality.
+*   AVOID defining custom equality for mutable classes.
+*   DON'T make the parameter to `==` nullable.
+
+---
+
+## Flutter Architecture Recommendations
+
+This page presents architecture best practices, why they matter, and
+whether we recommend them for your Flutter application.
+You should treat these recommendations as recommendations,
+and not steadfast rules, and you should
+adapt them to your app's unique requirements.
+
+The best practices on this page have a priority,
+which reflects how strongly the Flutter team recommends it.
+
+* **Strongly recommend:** You should always implement this recommendation if
+  you're starting to build a new application. You should strongly consider
+  refactoring an existing app to implement this practice unless doing so would
+  fundamentally clash with your current approach.
+* **Recommend**: This practice will likely improve your app.
+* **Conditional**: This practice can improve your app in certain circumstances.
+
+### Separation of concerns
+
+You should separate your app into a UI layer and a data layer. Within those layers, you should further separate logic into classes by responsibility.
+
+#### Use clearly defined data and UI layers.
+**Strongly recommend**
+
+Separation of concerns is the most important architectural principle.
+The data layer exposes application data to the rest of the app, and contains most of the business logic in your application.
+The UI layer displays application data and listens for user events from users. The UI layer contains separate classes for UI logic and widgets.
+
+#### Use the repository pattern in the data layer.
+**Strongly recommend**
+
+The repository pattern is a software design pattern that isolates the data access logic from the rest of the application.
+It creates an abstraction layer between the application's business logic and the underlying data storage mechanisms (databases, APIs, file systems, etc.).
+In practice, this means creating Repository classes and Service classes.
+
+#### Use ViewModels and Views in the UI layer. (MVVM)
+**Strongly recommend**
+
+Separation of concerns is the most important architectural principle.
+This particular separation makes your code much less error prone because your widgets remain "dumb".
+
+#### Use `ChangeNotifiers` and `Listenables` to handle widget updates.
+**Conditional**
+
+> There are many options to handle state-management, and ultimately the decision comes down to personal preference.
+
+The `ChangeNotifier` API is part of the Flutter SDK, and is a convenient way to have your widgets observe changes in your ViewModels.
+
+#### Do not put logic in widgets.
+**Strongly recommend**
+
+Logic should be encapsulated in methods on the ViewModel. The only logic a view should contain is:
+* Simple if-statements to show and hide widgets based on a flag or nullable field in the ViewModel
+* Animation logic that relies on the widget to calculate
+* Layout logic based on device information, like screen size or orientation.
+* Simple routing logic
+
+#### Use a domain layer.
+**Conditional**
+
+> Use in apps with complex logic requirements.
+
+A domain layer is only needed if your application has exceeding complex logic that crowds your ViewModels,
+or if you find yourself repeating logic in ViewModels.
+In very large apps, use-cases are useful, but in most apps they add unnecessary overhead.
+
+### Handling data
+
+Handling data with care makes your code easier to understand, less error prone, and
+prevents malformed or unexpected data from being created.
+
+#### Use unidirectional data flow.
+**Strongly recommend**
+
+Data updates should only flow from the data layer to the UI layer.
+Interactions in the UI layer are sent to the data layer where they're processed.
+
+#### Use `Commands` to handle events from user interaction.
+**Recommend**
+
+Commands prevent rendering errors in your app, and standardize how the UI layer sends events to the data layer.
+
+#### Use immutable data models.
+**Strongly recommend**
+
+Immutable data is crucial in ensuring that any necessary changes occur only in the proper place, usually the data or domain layer.
+Because immutable objects can't be modified after creation, you must create a new instance to reflect changes.
+This process prevents accidental updates in the UI layer and supports a clear, unidirectional data flow.
+
+#### Use freezed or built_value to generate immutable data models.
+**Recommend**
+
+You can use packages to help generate useful functionality in your data models, `freezed` or `built_value`.
+These can generate common model methods like JSON ser/des, deep equality checking and copy methods.
+These code generation packages can add significant build time to your applications if you have a lot of models.
+
+#### Create separate API models and domain models.
+**Conditional**
+
+> Use in large apps.
+
+Using separate models adds verbosity, but prevents complexity in ViewModels and use-cases.
+
+### App structure
+
+Well organized code benefits both the health of the app itself, and the team working on the code.
+
+#### Use dependency injection.
+**Strongly recommend**
+
+Dependency injection prevents your app from having globally accessible objects, which makes your code less error prone.
+We recommend you use the `provider` package to handle dependency injection.
+
+#### Use `go_router` for navigation.
+**Recommend**
+
+Go_router is the preferred way to write 90% of Flutter applications.
+There are some specific use-cases that go_router doesn't solve,
+in which case you can use the `Flutter Navigator API` directly or try other packages found on `pub.dev`.
+
+#### Use standardized naming conventions for classes, files and directories.
+**Recommend**
+
+We recommend naming classes for the architectural component they represent.
+For example, you may have the following classes:
+
+* HomeViewModel
+* HomeScreen
+* UserRepository
+* ClientApiService
+
+For clarity, we do not recommend using names that can be confused with objects from the Flutter SDK.
+For example, you should put your shared widgets in a directory called `ui/core/`,
+rather than a directory called `/widgets`.
+
+#### Use abstract repository classes
+**Strongly recommend**
+
+Repository classes are the sources of truth for all data in your app,
+and facilitate communication with external APIs.
+Creating abstract repository classes allows you to create different implementations,
+which can be used for different app environments, such as "development" and "staging".
+
+### Testing
+
+Good testing practices makes your app flexible.
+It also makes it straightforward and low risk to add new logic and new UI.
+
+#### Test architectural components separately, and together.
+**Strongly recommend**
+
+* Write unit tests for every service, repository and ViewModel class. These tests should test the logic of every method individually.
+* Write widget tests for views. Testing routing and dependency injection are particularly important.
+
+#### Make fakes for testing (and write code that takes advantage of fakes.)
+**Strongly recommend**
+
+Fakes aren't concerned with the inner workings of any given method as much
+as they're concerned with inputs and outputs. If you have this in mind while writing application code,
+you're forced to write modular, lightweight functions and classes with well defined inputs and outputs.

package/assets/instructions/self-explanatory-code-commenting/self-explanatory-code-commenting.instructions.md

@@ -1,5 +1,5 @@
 ---
-description: 'Guidelines for GitHub Copilot to write comments to achieve self-explanatory code with less comments. Examples are in JavaScript but it should work on any language that has comments.'
+description: 'Conventions for writing and reviewing code comments: prefer self-explanatory code, comment WHY not WHAT, and avoid obvious, redundant, outdated, dead-code, or decorative comments. Language-agnostic (examples in JavaScript).'
 applyTo: '**/*.py, **/*.js, **/*.ts, **/*.java, **/*.cs, **/*.cpp, **/*.go, **/*.rb'
 ---
 

package/assets/skills/vesper-catalog-editing-safety/SKILL.md

@@ -50,7 +50,7 @@ Most edits hit exactly one checklist. If unsure, read Checklist D first — it's
 | `assets/skills/<name>/`                                | `assets/claude/skills/<name>/` (near-verbatim)             |
 | `assets/instructions/<name>/<name>.instructions.md`    | `assets/claude/skills/<name>/SKILL.md` (new skill)         |
 
-**That table is the entire remit.** `assets/core/` is never read by this script, at all — by design, not by omission (the script's own comment: *"the 12 core agents + the orchestrator playbook are hand-authored separately in assets/claude/ and are never touched by this script"*).
+**That table is the entire remit.** `assets/core/` is never read by this script, at all — by design, not by omission (the script's own comment: *"the 12 core agents + the orchestrator playbook are hand-authored separately in assets/claude/ and are never touched by this script"*). Neither is `assets/vendor/` — reviewed, pinned third-party snapshots (e.g. `assets/vendor/i-have-adhd/`, `assets/vendor/ponytail/`) live outside the four scan roots and the three write targets on purpose. A vendored asset is edited only by re-vendoring (a reviewed PR bumping its pinned SHA), never through the generator; do not move it under `assets/skills/` or `assets/claude/` to "make the generator pick it up."
 
 - [ ] Is the file under `assets/core/`?
   - **Yes** → There is no generator step for you. Hand-edit **both** the Copilot source (`assets/core/{agents,skills,instructions,prompts}/...`) **and** its `assets/claude/` twin. Before editing the twin, check it for **Claude-only content that doesn't exist in the Copilot source** and preserve it — e.g. `assets/claude/skills/agent-fundamentals/SKILL.md`'s `## Direct Collaboration (Nested Delegation)` section has no Copilot-side counterpart. Never diff the two sides expecting them to match; they're allowed to diverge by design.
@@ -130,6 +130,7 @@ If your edit changes *where* either target looks, or *what* gets written into th
 
 - `assets/` — the Copilot-format source catalog (`core/` hand-authored default install; `agents/`, `skills/`, `instructions/`, `prompts/` extended/opt-in)
 - `assets/claude/` — generated Claude-target twin of the extended tree, plus hand-maintained Claude-only twins of `assets/core/`
+- `assets/vendor/` — reviewed, pinned snapshots of third-party assets (e.g. `assets/vendor/i-have-adhd/`, `assets/vendor/ponytail/`); a generator-excluded category, installed project-level at init time, never scanned/written by the generator and never a scan root
 - `scripts/generate-claude-assets.js` — the extended-tree generator; never touches `assets/core/`
 - `src/cli/utils/settings.js` — `VSCODE_SETTINGS_MAP`, `getCoreAssetEntries()`, `VESPER_GIT_PATHS`
 - `src/cli/utils/claude-target.js` — `getClaudePluginDir()`, `installClaudePluginCatalog()`, `ensureClaudeMdGuardBlock()`, `VESPER_REINJECT_COMMAND`

package/assets/vendor/i-have-adhd/LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2026 Ayoub Ghriss
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

package/assets/vendor/i-have-adhd/SKILL.md

@@ -0,0 +1,142 @@
+---
+name: i-have-adhd
+description: 'Shape output for a reader with ADHD: lead with the next action, number multi-step work, restate state across turns, suppress tangents, give specific time estimates, make wins visible. Invoke with /i-have-adhd; stays on until "stop adhd mode".'
+disable-model-invocation: true
+license: MIT
+metadata:
+  hermes:
+    tags: [ADHD, Output Style, Productivity, Formatting]
+    category: productivity
+    related_skills: []
+---
+
+# i-have-adhd
+
+The reader has ADHD. Output is not just brief. It is shaped so an ADHD brain can act on it.
+
+## Persistence
+
+These rules apply to every response for the rest of the session, not only this one. They do not expire after a few turns and they do not lapse when the topic changes. If you are unsure whether they still apply, they do.
+
+Turn them off only when the reader says "stop adhd mode" or "normal mode". Confirm in one line, then return to your default style.
+
+## What ADHD changes about reading
+
+Five facts drive every rule below:
+
+1. Working memory is small. Anything not on screen is forgotten. Do not ask the reader to "keep in mind X."
+2. Knowing the answer is not doing the answer. The friction between "got it" and "done it" is where work dies.
+3. Starting is the hardest step. The first action must be obvious, small, and doable now.
+4. Time estimates feel uniform. "A bit of work" and "a few hours" register the same. Vague estimates fail.
+5. Dopamine is scarce. Visible progress matters. Buried wins do not register.
+
+## Rules
+
+### 1. Lead with the next action
+
+The first line is something the reader can do. Not context. Not a plan. The action.
+
+Bad: "Let's think about this. Your auth flow has a few moving pieces..."
+Good: "Run `npm install jsonwebtoken`, then edit `src/auth.ts:42`."
+
+If the answer is a command, path, or snippet, it goes first. Prose comes after, if at all.
+
+### 2. Number multi-step tasks
+
+If the work takes more than one step, write a numbered list. Each step is one bounded action. No step contains "and then" twice.
+
+Use the fewest steps that still work. Cut any step the reader does not need, and fold trivial steps into the one before. A short path finished beats a complete path abandoned.
+
+Bad: "First open the file, find the function, swap it out, then run the tests."
+
+Good:
+```
+1. Open `src/auth.ts`
+2. Replace `verifyToken` (lines 42 to 58) with the snippet below
+3. Run `npm test -- auth.spec.ts`
+```
+
+### 3. End with one concrete next action
+
+If anything is left open, name ONE thing the reader can do in under two minutes. Even "open the file" counts.
+
+Bad: "Hope that helps. Let me know if you want to dig deeper."
+Good: "Next: run `npm test` and paste the first failing line."
+
+### 4. Suppress tangents
+
+If a second issue exists, finish the first, then offer the second as a separate question.
+
+Bad: "Here's the fix. By the way, your dependency is also stale, and your README is out of date, and..."
+Good: "Here's the fix. Separately: there is also a stale dependency. Want me to handle that next?"
+
+A question that comes up mid-work is not a tangent: answer it yourself if you can and fold the result in. If it still needs the reader, surface it once, at the end.
+
+### 5. Restate state every turn
+
+The reader cannot hold "we are on step 3 of 5" between messages. Restate it.
+
+Bad: "Done. Ready for the next part?"
+Good: "Step 3 of 5 done: schema updated. Next: backfill the new column. Run the script?"
+
+If the harness has a task or plan tool, use it for multi-step work: one item per step, one in progress at a time. The checklist does the restating; do not also narrate the full plan as prose.
+
+### 6. Give specific time estimates
+
+Vague estimates fail. Ballpark in concrete units.
+
+Bad: "This will take some work."
+Good: "About 15 minutes if tests already cover this. An afternoon if not."
+
+### 7. Make completed work visible
+
+Show what now works, in concrete terms. Do not bury wins in a recap.
+
+Bad: "I've made some changes to the auth flow. Among other things..."
+Good: "Login now works with magic links. Try: `npm run dev`, open `/login`."
+
+### 8. Matter-of-fact tone for errors
+
+Never use "Uh oh," "Oh no," or "There seems to be a problem." State cause and fix.
+
+Bad: "Uh oh, the test is failing. There seems to be an issue..."
+Good: "Test fails at `auth.spec.ts:42`: expected 200, got 401. Cause: missing auth header. Fix: add `Authorization: Bearer ${token}` to the request."
+
+### 9. Cap lists at 5 items
+
+If a list grows past five, split into "do now" vs "later," or "must" vs "nice to have." Five items ranked beats ten unranked.
+
+### 10. No preamble, no recap, no closing pleasantries
+
+Forbidden openers: "Great question," "Let me...", "I'll...", "Sure!", "Looking at your...", "To answer your question..."
+
+Forbidden recaps after a completed task: "I've now done X, Y, and Z, which means..."
+
+Forbidden closers: "Let me know if you need anything else," "Hope this helps," "Happy to clarify," "Feel free to ask."
+
+Start with the answer. End when the answer is done.
+
+## When to break the rules
+
+Override the defaults when:
+
+1. User asks to "explain" or "walk me through." Explain fully. Still no preamble, still no closer, but the body runs as long as the topic needs. Add headers so the reader can skim back.
+2. Destructive action ahead (`rm -rf`, force push, schema migration, dropping a table). Confirm before acting. Safety wins over brevity.
+3. Debug spiral. If the last three turns have been "still broken," stop iterating on code. Name the assumption that might be wrong. Ask one diagnostic question.
+4. Real ambiguity in the request. One short clarifying question beats guessing and rewriting.
+5. A rule fights the task. When a rule would delete the answer itself, the task wins; the shape stays. Example: "what are my options" gets 2 to 4 ranked options with one-line trade-offs, recommendation first, not one path. The options are the answer.
+6. A rule fights the harness. Inside an agent harness, the system prompt outranks this skill: announce a tool call when the harness requires it, do the work instead of asking "want me to," point time estimates at whoever executes the steps. Same principle as 5: the constraint wins, the shape stays.
+
+## Pre-send check
+
+Before sending, delete:
+
+1. The first sentence if it announces what you are about to do.
+2. The last sentence if it asks "anything else?" or recaps what just happened.
+3. Any "by the way" sidebar.
+4. Any hedging adverb adding no information ("perhaps," "might," "could possibly"). Keep a hedge that carries real uncertainty; deleting it manufactures confidence.
+5. Any idiom or figurative phrase ("circle back," "get the ball rolling," "on the same page"). Replace with the literal action.
+
+Then verify: if the reader reads only the first line and the last line, do they know (a) what to do next, and (b) what just happened?
+
+If yes, send.