Struct LoopSkill 

pub struct LoopSkill {
    pub name: String,
    pub description: Option<String>,
    pub version: Option<String>,
    pub scope: SkillScope,
    pub dir: PathBuf,
    pub manifest: PathBuf,
    pub model_invocable: bool,
    pub allowed_tools: Vec<String>,
    pub argument_names: Vec<String>,
    pub argument_hint: Option<String>,
}

One SKILL.md package the supercode loop itself will load.

Fields

name: String

The name the loop invokes it by — frontmatter name (else the directory name), qualified plugin:skill / dir:skill where the source harness qualifies it.

description: Option<String>

Frontmatter description, one line. The INDEX line’s whole payload; a skill without one is still invocable, just undescribed.

version: Option<String>

Frontmatter version.

scope: SkillScope

Precedence class of the root it came from.

dir: PathBuf

The skill’s own directory.

manifest: PathBuf

<dir>/SKILL.md — where the BODY lives, read only on invocation.

model_invocable: bool

Whether the model may see it in the prompt index at all. false for enabled: false / disable-model-invocation: true frontmatter: the skill stays user-invocable by name, it is simply not advertised (cc§7 “Invocation control”, pi§2).

allowed_tools: Vec<String>

BP-5 (cc§7 “Invocation control”: “pre-approved tools while active”): the package’s own allowed-tools frontmatter, verbatim. Its only consumer is ShellInjection::expand, where it pre-approves this body’s OWN !`cmd` commands and nothing else — it never widens what the model’s tool calls are allowed to do.

argument_names: Vec<String>

BP-5 (cc§7 “Skill frontmatter”: arguments (named positional)): the package’s ARGUMENT SCHEMA — the names its body substitutes as $name, in positional order. Empty when the body only uses $ARGUMENTS/$1..$9.

argument_hint: Option<String>

BP-5 (cc§7 argument-hint): the one-line usage hint shown beside this package in the prompt index, so a model calling it by name knows what the trailing text should be.

Implementations

impl LoopSkill

pub fn index_line(&self) -> String

The prompt-index line: name and description only — never the body.

pub fn body(&self, arguments: &str) -> Result<String, Error>

Read this skill’s BODY (everything after the frontmatter fence), substituting $ARGUMENTS / $1..$9 with the invocation’s trailing text. This is the ONLY function that spends a body’s tokens; nothing on the discovery path reads past the frontmatter.

pub fn body_with_shell( &self, arguments: &str, shell: &ShellInjection, ) -> Result<String, Error>

BP-5: Self::body, then the !`cmd` expansion shell authorizes (cc§7 “Dynamic context injection”). This is the door every invocation surface uses — the skill tool, /name, /skill:name and $slug — so one body cannot mean two things depending on which door loaded it. With shell injection off (the default) this is exactly Self::body.

Trait Implementations

impl Clone for LoopSkill

fn clone(&self) -> LoopSkill

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for LoopSkill

fn fmt(&self, f: &mut Formatter<'_>) -> Result<(), Error>

Formats the value using the given formatter.

impl<'de> Deserialize<'de> for LoopSkill

fn deserialize<__D>( __deserializer: __D, ) -> Result<LoopSkill, <__D as Deserializer<'de>>::Error>

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl Eq for LoopSkill

impl PartialEq for LoopSkill

fn eq(&self, other: &LoopSkill) -> bool

Equality operator ==.

fn ne(&self, other: &Rhs) -> bool

Inequality operator !=.

impl Serialize for LoopSkill

fn serialize<__S>( &self, __serializer: __S, ) -> Result<<__S as Serializer>::Ok, <__S as Serializer>::Error>

where __S: Serializer,

Serialize this value into the given Serde serializer.

impl StructuralPartialEq for LoopSkill