SkillGraphBuilder
Interface: SkillGraphBuilder
Defined in: src/lib/injection-engine/skillGraph.ts:278
Methods
build()
build(
opts?):SkillGraph
Defined in: src/lib/injection-engine/skillGraph.ts:321
Parameters
opts?
Returns
entry()
entry(
skill,opts?):SkillGraphBuilder
Defined in: src/lib/injection-engine/skillGraph.ts:280
Mark a skill as reachable at turn start (optionally intent-conditional).
Parameters
skill
opts?
Returns
SkillGraphBuilder
entryBy()
entryBy(
scorer):SkillGraphBuilder
Defined in: src/lib/injection-engine/skillGraph.ts:297
Pick the STARTING entry with a pluggable scorer STRATEGY — keywordScorer()
(no dependency, word overlap), embeddingScorer(embedder) (semantic), or your
own EntryScorer. The agent's PickEntry stage runs it ONCE per turn off the
hot loop and starts the cursor at the winner. Like .entryByRead(), this makes
the entries EXCLUSIVE (only the chosen one loads, token-efficient). The surfaced
relevance % powers the "Why this skill?" panel. Flat graphs only (a decision
tree() already routes by predicate). Mutually exclusive with .entryByRead().
Parameters
scorer
Returns
SkillGraphBuilder
entryByRead()
entryByRead():
SkillGraphBuilder
Defined in: src/lib/injection-engine/skillGraph.ts:320
Let the LLM pick the STARTING entry by reading the menu — no embedder, no extra
model call. Like .entryByRelevance(), the entries become EXCLUSIVE (only the
chosen one loads, token-efficient), but the choice is the model's: on the first
turn no entry auto-loads, the agent is offered the entries via read_skill, and
its pick becomes the cursor. Use this when you have NO embedder (or embeddings
route poorly for your domain) — the agent's own LLM understands the request.
Flat graphs only; mutually exclusive with .entryByRelevance().
Caveat: prefer UNCONDITIONAL entries here. A when-gated entry may still be
offered in the read_skill menu (the cold-start gate can't evaluate when), but
if the model picks it while its when is false it won't load — the turn wastes
an iteration with no skill. For intent-gating, use .entryByRelevance() or plain
.entry(s, { when }) (v1 always-on) instead.
Returns
SkillGraphBuilder
entryByRelevance()
entryByRelevance(
embedder):SkillGraphBuilder
Defined in: src/lib/injection-engine/skillGraph.ts:304
Sugar for .entryBy(embeddingScorer(embedder)) — pick the starting entry by
SEMANTIC relevance (embed the message + each entry's description, cosine-score,
softmax → best match). LLM-free (an embedder, no extra model call), reproducible.
For a no-embedder router, use .entryBy(keywordScorer()).
Parameters
embedder
Returns
SkillGraphBuilder
route()
route(
from,to,opts?):SkillGraphBuilder
Defined in: src/lib/injection-engine/skillGraph.ts:282
Declare an edge: after from's work, to activates when the edge fires.
Parameters
from
to
opts?
Returns
SkillGraphBuilder
tree()
tree(
root,opts?):SkillGraphBuilder
Defined in: src/lib/injection-engine/skillGraph.ts:287
Declare a decision TREE (v3): predicate nodes → skill leaves. Compiles each
leaf to a path-conjunction trigger; renders as diamonds → boxes. By default
each leaf is tool-scoped (autoActivate: 'currentSkill') so only the routed
skill's tools reach the LLM — opt out with { scopeTools: false }.
Parameters
root
opts?
Returns
SkillGraphBuilder