Interface: SkillGraph

Defined in: src/lib/injection-engine/skillGraph.ts:227

Properties

edges

readonly edges: readonly SkillEdge[]

Defined in: src/lib/injection-engine/skillGraph.ts:232

The declared edges (for tooling, overlays, tests).

---

nodes

readonly nodes: readonly SkillNode[]

Defined in: src/lib/injection-engine/skillGraph.ts:235

Drawn nodes: skill boxes for the flat entry/route model; predicate diamonds

• skill leaves for a decision tree. Always present.

---

skills

readonly skills: readonly Injection[]

Defined in: src/lib/injection-engine/skillGraph.ts:230

Skills with graph-derived triggers — feed to the Agent (.skillGraph() or .skills({ list: () => graph.skills })).

Methods

checkup()

checkup(): GraphCheckup

Defined in: src/lib/injection-engine/skillGraph.ts:275

Build-time check-up — inspect the declared graph for wiring mistakes (a skill nobody can reach, an edge to an unknown skill, two un-prioritized edges from one skill, no entry, a self-loop). Pure + side-effect-free; call it whenever. ok is false iff there's an error-level problem (unknown-skill / no-entry).

Returns

GraphCheckup

---

nextSkill()

nextSkill(ctx): string | undefined

Defined in: src/lib/injection-engine/skillGraph.ts:249

The CURSOR resolver — given an iteration context, where is the graph next? Returns the skill id the graph should be in after this iteration: • cold start (ctx.currentSkillId unset) → the first matching entry; • a from-gated route whose predicate matches ctx.lastToolResult → its target; • otherwise the current cursor unchanged (sticky stay). Pure + deterministic — the single source of truth shared by the compiled route triggers and the agent loop's cursor-update stage, so the two can never disagree. Flat entry/route graphs only; a decision tree() routes per-iteration by predicate (no cursor) and returns the unchanged ctx.currentSkillId.

Parameters

ctx

InjectionContext

Returns

string | undefined

---

reachableSkills()

reachableSkills(currentSkillId?): readonly string[]

Defined in: src/lib/injection-engine/skillGraph.ts:260

The REACHABLE set — which skills the model may read_skill-jump to from the current cursor. The agent's runtime gate rejects any read_skill('id') whose id is not in this set (so the model can't leave the graph mid-run). • cold start (currentSkillId undefined) → the entry skills; • otherwise → the current skill's direct successors ∪ the entry skills, minus the current skill itself (deliberate "stay" is the no-tool-call ReAct stop). Pure + deterministic. A decision tree() has no cursor, so it returns ALL leaf skills — read_skill stays a full escape hatch there.

Parameters

currentSkillId?

string

Returns

readonly string[]

---

scoreEntries()?

optional scoreEntries(ctx, signal?): Promise<EntryScoring>

Defined in: src/lib/injection-engine/skillGraph.ts:268

Score the entry candidates by relevance to the user's message — present ONLY when the graph was built with .entryByRelevance(embedder). Embeds ctx.userMessage and each when-passing entry's description, cosine-scores them, and softmaxes into a relevance share. The agent's PickEntry stage uses chosen as the starting cursor (LLM-free, off the hot loop). Flat graphs only.

Parameters

ctx

InjectionContext

signal?

AbortSignal

Returns

Promise<EntryScoring>

---

toMermaid()

toMermaid(): string

Defined in: src/lib/injection-engine/skillGraph.ts:237

A Mermaid flowchart of the declared graph — declared === drawn.

Returns

string