Struct SkillFrontmatter 

pub struct SkillFrontmatter {
    pub name: String,
    pub description: String,
    pub applies_to: Option<HashMap<String, String>>,
    pub references_tools: Vec<String>,
    pub references_arguments: Vec<String>,
    pub references_properties: Vec<String>,
    pub auto_inject_hint: bool,
    pub applies_when: Option<AppliesWhen>,
}

Parsed YAML frontmatter of a SKILL.md file.

Phase 1b stores all declared fields as raw values. Phase 1f / 2a will add validation (applies_to semver checks, references_tools against the active tool catalogue, references_arguments against each tool’s input schema). For now: parse and preserve; the lint step in Registry::finalise() walks these and surfaces issues as log warnings.

Fields

name: String

Skill name. Must match the lookup key used in prompts/get. Required; empty after deserialization triggers a clear SkillError::MissingRequiredField rather than a generic YAML parse failure.

description: String

One-line description shown in prompts/list. Required — the agent uses this to decide whether to load the full body.

applies_to: Option<HashMap<String, String>>

Version constraints. Parsed lazily — Phase 1b stores raw values, Phase 1f adds semver validation.

references_tools: Vec<String>

Tools this skill teaches or references in prose. Used for auto-inject discoverability hints (Phase 1c) and staleness detection (Phase 1f).

references_arguments: Vec<String>

Specific tool argument names referenced in the skill body (e.g. "cypher_query.format"). Lint warns when references don’t match the tool’s actual input schema.

references_properties: Vec<String>

Graph properties / domain-specific references the skill calls out (e.g. "Function.module"). For domain skill-packs to declare their domain assumptions. The framework can’t validate these statically; they’re documentation-grade metadata.

auto_inject_hint: bool

When true (the default) AND the skill’s name matches a registered MCP tool, the framework injects a “see prompts/get <name> for full methodology” pointer into the tool’s description. Phase 1c wires this up.

applies_when: Option<AppliesWhen>

applies_when: predicate set. Bounded — not a DSL. All populated fields must evaluate true (AND semantics) for the skill to surface in prompts/list and prompts/get. The framework dispatches tool_registered and extension_enabled itself; domain predicates (graph_has_node_type, graph_has_property) are evaluated via the optional SkillPredicateEvaluator registered on the Registry.

None (the default) means “always active” — the skill applies regardless of runtime state.

Trait Implementations

impl Clone for SkillFrontmatter

fn clone(&self) -> SkillFrontmatter

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for SkillFrontmatter

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

Formats the value using the given formatter.

impl Default for SkillFrontmatter

fn default() -> SkillFrontmatter

Returns the “default value” for a type.

impl<'de> Deserialize<'de> for SkillFrontmatter

fn deserialize<__D>(__deserializer: __D) -> Result<Self, __D::Error>

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.