grounded-forge

Source-grounded answer to a question via shape-aware retrieval against the dist-only corpus: classifies the query as a named lookup, a diagnostic, or a synthesis, then runs the right protocol (direct route via slug, task-index → distillations, or three-pass index → distillations). Stops on sub-claim coverage, not on counts.

Projects one existing verified reference onto one existing task axis, gated by a per-distillation applicability check. Produces one distillation file at `corpus.commons/{corpus}/distillations/{task}/{slug}-{task}.md`, optionally with lens-applied sections or per-lens variants. Invoked by `ingesting-resources` Pass G during new-source ingestion and by `creating-applications` during per-distillation orchestration when assembling a new application.

Identifies sources in the corpus that match a query topic, situation, or concept. Returns a list with one-line justifications for each candidate, with paths to the relevant distillations.

Run a 10–15 minute structured interview with the operator that surfaces register, current circumstance, working style, relational stance, and values, then write a self-lens spec at `{corpus}/lenses/{operator-slug}-self.md`. The lens is a named-real-person lens (per `creating-lenses`) where the named person is the operator. Delta-mode by default: skips territory already covered by global or project CLAUDE.md. Voice-mode auto-detected when the `voicemode` skill is available. The lens spec is ready for `creating-distillations` Pass G.

Runs the 9-pass ingestion protocol against a single source dropped in {corpus-root}/sources/ingest/, producing a .source.md sidecar, the converted markdown, a deep reference, a light reference, one distillation per applicable task axis, and index and image updates. On success, the input source is removed from sources/ingest/.

Audit an essay or draft for under-attribution against the library's curated corpus: surfaces the one or two closest neighbours whose work overlaps with the author's distinctive synthesis, classifies the overlap, and recommends specific provenance fixes. Not a fact-check; a build/contribute-boundary check.

Design a lens: the per-distillation modifier described in `docs/reference/vocabulary.md` that shapes a distillation through a perspective, role, or methodology stance. The skill is a single-process dialogue; the output is a lens spec at `corpus.commons/{corpus}/lenses/{lens-slug}.md`, ready to feed Pass G of `creating-distillations` when the operator decides which distillations the lens applies to.

Discovers and triages candidate sources before ingestion. Produces a triage report with licence, distribution scope, format, viability verdict, and gap-fit notes: three buckets (proceed, sample first, skip). Each `proceed` entry is shaped as ready input for `ingesting-resources`.

The single entry point for creating a new application: a compiled distribution built from a corpus subset projected onto a task axis (with optional lenses + practitioner-role agents). Run this when a user says "create an app for X" or similar. The skill first ensures a task spec exists (calling `creating-tasks` internally if one does not, so the user sees a continuous experience rather than a bounce), then scaffolds the distillation directory, distillation index, build profile, CLAUDE.md template, and agent files, then orchestrates per-distillation distillation by calling `creating-distillations` for each applicable source.

Scaffold a new corpus under corpus.commons/ or corpus.local/: walk the operator through tier, naming, and task axes, then hand off to scripts/create-corpus.py for the filesystem work. The output is a self-contained corpus tree (sources, references, distillations, lenses, dist, distros) ready to receive material into sources/ingest/.

Scope a new task axis for the matrix: the column-direction in the reference × task projection. The skill is a single-process dialogue grounded in Jobs-to-be-Done discipline; the output is a task spec at `corpus.commons/{corpus}/tasks/{task-slug}.md`, ready to feed `creating-applications` for assembly into a compiled distribution.

Classifies images extracted from an already-ingested source as substantive (diagrams, charts, frameworks) or decorative (photos, covers, logos), and appends substantive entries to the corpus image index.