Struct SkillManager 

pub struct SkillManager { /* private fields */ }

Implementations

impl SkillManager

pub fn from_skills_dir(skills_dir: &str) -> Self

pub fn from_skills_and_runtime(skills_dir: &str, runtime_dir: &str) -> Self

pub fn new(data_dir: &str) -> Self

pub fn with_clawhub_verification(self, lockfile: PathBuf, block: bool) -> Self

Enable load-time verification of ClawHub-managed skill trees against lockfile. With block true (config clawhub_verify_on_load: block, the default) a mismatching skill becomes unavailable; with block false it only logs.

pub fn with_config_verification(self, config: &Config) -> Self

Apply the configured ClawHub load-time verification policy.

pub fn discover_skills(&self) -> Vec<SkillMetadata>

Discover all skills that are available on the current platform and satisfy dependency checks.

pub fn discover_skills_with_status( &self, include_unavailable: bool, ) -> Vec<SkillAvailability>

Discover skills with availability diagnostics.

pub fn reload(&self) -> Vec<SkillMetadata>

Reload skills from disk (live reload)

pub fn has_skill(&self, name: &str) -> bool

pub fn set_enabled(&self, name: &str, enabled: bool) -> Result<(), String>

pub fn load_skill(&self, name: &str) -> Option<(SkillMetadata, String)>

Load a skill by name if it is available on the current platform.

pub fn load_skill_checked( &self, name: &str, ) -> Result<(SkillMetadata, String), String>

Load a skill with availability diagnostics.

pub fn build_skills_catalog(&self) -> String

Build a compact skills catalog for the system prompt. Returns empty string if no skills are available.

pub fn build_skills_hot_bodies_for_query( &self, query: &str, top_k: usize, ) -> String

Render the query-matched skills with their full instruction bodies inlined, as a self-contained <turn_skills> block.

This is the query-scoped half of the skills context: it changes with every user question, so it belongs in the per-turn context block rather than in the cached system-prompt prefix. Returns an empty string when the query matched nothing.

pub fn build_skills_catalog_for_query( &self, query: &str, top_k: usize, ) -> String

Build a query-aware skills catalog: inline the full body of the top-top_k skills whose descriptions overlap the query, and fall back to a compact name: description listing for the rest.

Tradeoff vs build_skills_catalog: spending a bigger token slice on the most-relevant skills (so the agent has the procedural knowledge inline and doesn’t need an extra activate_skill round-trip) while keeping the long tail cheap.

top_k = 0 falls back to [build_skills_catalog] verbatim. An empty query also falls back — without a query the relevance score is meaningless.

pub fn build_skills_catalog_for_names_and_query( &self, names: &[String], _query: &str, _top_k: usize, ) -> String

Build a catalog restricted to an Agent Profile’s explicit skill selection.

Unlike the global query-aware catalog, explicitly assigned skills are all included with their instructions (subject to the same per-skill body cap). Missing, disabled, or incompatible skills are represented as unavailable instead of silently enabling a different skill.

pub fn list_skills_formatted(&self) -> String

Build a user-facing formatted list of available skills.

pub fn list_skills_formatted_all(&self) -> String

Build a user-facing list, optionally including unavailable skills and reasons.

pub fn skills_dir(&self) -> &PathBuf

pub fn state_file_path(&self) -> Option<&Path>