Struct SkillRecord 

pub struct SkillRecord {

    pub id: String,
    pub name: Option<String>,
    pub layers: Vec<usize>,
    pub selection: Option<SelectionDescriptor>,
    pub input_mask_task: Option<String>,
    pub quality: Option<Value>,
    pub base_dir_hash: Option<String>,
    pub base_arch: Option<String>,
    pub task: Option<String>,
    pub provenance: Option<Value>,
    pub kind: Option<String>,
    pub experts: Option<ExpertAppend>,
    pub lookup: Option<LookupInfo>,
    pub overrides: Vec<SkillOverride>,
    pub bound: Option<SkillBound>,
    pub state_effect: Option<StateEffect>,
    pub status: Option<String>,
    pub gate: Option<Value>,
    pub prompt_contract: Option<String>,
    pub origin: Option<Value>,
}

One skill of the swarm (spec §9; Patent 15 per-skill record).

Fields

id: String

name: Option<String>

layers: Vec<usize>

Layers this skill specializes (a proper subset).

selection: Option<SelectionDescriptor>

Selection descriptor for recon-argmin routing (208c, P1): per-skill affine subspace over φ(x) = mean-pooled hidden state.

input_mask_task: Option<String>

Optional input-mask task name (208b), applied with the skill.

quality: Option<Value>

Measured quality (claim 16): overlaid vs backbone, held-out.

base_dir_hash: Option<String>

hex hash64 of the BASE model’s tensor directory this skill was cut against. skill apply refuses a base whose directory hash differs: a skill is a delta against exact bytes, not an architecture.

base_arch: Option<String>

The base’s architecture name — the cheap human-readable identity check next to the exact one above.

task: Option<String>

Mask-catalog task this skill activates once applied.

provenance: Option<Value>

Free-form provenance: corpus, steps, recipe, who baked it.

kind: Option<String>

Record kind: “ffn_replace” and “expert_append” (implemented). Reserved and refused until implemented: “anchor_sinks”, “mask_only”, “fcd_vector”. None = a v1 record (unchanged semantics).

experts: Option<ExpertAppend>

expert_append parameters (spec §9.5.1): count new experts in each of layers, their descriptor rank and the shell_quantile the trainer used. Required for that kind, refused on any other.

lookup: Option<LookupInfo>

lookup parameters (spec §9.5.2): the explicit key → card table under skill.{id}.lookup.* — entries, keys, key normalisation, languages, fields. Required for that kind, refused on any other.

overrides: Vec<SkillOverride>

Base tensors this record replaces, with the hash of the base entry it was trained against. Exactly one per skill.{id}.X tensor.

bound: Option<SkillBound>

The genome (id, generation, f32-master trunk hash) the record is bound to.

state_effect: Option<StateEffect>

What switching this skill does to the sequence state (computed by the writer from arch.layer_types).

status: Option<String>

“quarantine” | “active” | “stale_regate” | “retired”.

gate: Option<Value>

Measured gate; auto-routing requires gate.status == "measured" and status == "active".

prompt_contract: Option<String>

Prompt template the skill was trained under (e.g. “cmf-im-v1”).

origin: Option<Value>

{trigger, dataset_sha256[], recipe, …}.

Implementations

impl SkillRecord

pub fn is_v2(&self) -> bool

A v2 record (kind set) — validated by the SKILLS_V2 rules.

pub fn is_auto_routable(&self) -> bool

May the router-v2 policy auto-route to this skill? Requires a v2 record with a selection descriptor, status == "active" and a measured gate (gate.status == "measured").

Trait Implementations

impl Clone for SkillRecord

fn clone(&self) -> Self

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for SkillRecord

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

Formats the value using the given formatter.

impl Default for SkillRecord

fn default() -> Self

Returns the “default value” for a type.

impl<'de> Deserialize<'de> for SkillRecord

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

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl Serialize for SkillRecord

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

where __S: Serializer,

Serialize this value into the given Serde serializer.