Summary of Changes

Hello, I'm Gemini Code Assist¹! I'm currently reviewing this pull request and will post my feedback shortly. In the meantime, here's a summary to help you and other reviewers quickly get up to speed!

This pull request enhances the tool argument parsing logic to robustly handle scoped npm package names. The primary goal is to prevent application panics and provide clearer error feedback to users when such package names are provided without an explicit backend prefix, ensuring a more stable and user-friendly experience.

Highlights

• Panic Fix: Resolved a panic that occurred when using scoped npm package names (e.g., @anthropic-ai/claude-code) without the npm: backend prefix.
• Parsing Logic Improvement: The parse_input function now correctly recognizes a leading @ as part of the package name for scoped packages, rather than a version separator.
• User Experience: Users will now receive a proper error message instead of encountering a panic with a Pre-condition of new violated: ! short.is_empty() error.
• Test Coverage: Added new unit tests specifically for scoped package parsing, covering cases like @scope/pkg, @scope/pkg@version, and @scope/pkg@.

Greptile Summary

This PR fixes a panic that occurred when a scoped npm package name (e.g., @anthropic-ai/claude-code) was passed to mise without the npm: backend prefix. The root cause was that parse_input split on the first @ and returned an empty string as the backend name, which then violated the #[requires(!short.is_empty())] precondition in BackendArg::new.

Changes:

• A new guard in parse_input detects when left is empty after the initial split_once('@') (indicating a scoped package starting with @), then searches right for a second @ to separate the package name from an optional version specifier.
• The byte-slice arithmetic (&s[..name.len() + 1]) correctly reconstructs the leading @ plus the package name.
• Five new test cases in test_tool_arg_parse_input cover: no-version, two different scoped names, versioned (@latest, @1.0.0), and trailing-@ forms.

Confidence Score: 5/5

• This PR is safe to merge — it is a tightly scoped bug fix with no regressions and good test coverage.
• The change is confined to a single function (parse_input) and is purely additive — it inserts a guard branch before any existing logic. The byte-slice arithmetic is correct: right is s[1..], so name.len() + 1 is a valid UTF-8 boundary in s (since @ is single-byte ASCII and name is a substring of right). The five new unit tests validate all edge cases (no version, semver, dist-tag, empty trailing @), and the existing tests confirm no regressions in backend-prefix or simple tool@version paths.
• No files require special attention.

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart TD
    A["parse_input(s)"] --> B["split_once('@') on s"]
    B -->|"No '@' found"| C["Return (s, None)"]
    B -->|"left='', right=remainder"| D{"left.is_empty()?"}
    D -->|"Yes – scoped package"| E["split_once('@') on right"]
    E -->|"Found second '@'"| F["Return (&s[..name.len()+1], version)"]
    E -->|"No second '@'"| G["Return (s, None)"]
    D -->|"No"| H{"left.ends_with(':')?"}
    H -->|"Yes – backend prefix"| I["split_once('@') on right for version"]
    I -->|"Found '@'"| J["Return (backend_tool, version)"]
    I -->|"Not found"| K["Return (s, None)"]
    H -->|"No – simple tool@version"| L["Return (left, right_or_none)"]

_{Last reviewed commit: ac83031}

bugbot run

Hyperfine Performance

mise x -- echo

mise env

mise hook-env

mise ls

xtasks/test/perf