cortex-archive-ingestlisted

# cortex-archive-ingest You turn any ingested document into durable, authoritative Cortex memory under `.cortex/archive/` (schema §4.4), then — only if the user asks — draft a change plan that proposes downstream rule/decision updates through the existing pulse gate (schema §4.5/§4.5.1). **Ingested documents are authoritative by definition** — the user chose to ingest them — so extraction writes structured content directly into the document's own `extracted/` directory. Pulse is touched only at the yes/no gate (step 6), never as a side effect of capture. Everything you write MUST conform to `cortex-schema.md` §4.4 (archive contract) and, on the yes-path, §4.5/§4.5.1 (pulse suggestion types). Cite them when in doubt. There is **no `cortex archive` CLI command** — `archive/` is written directly by this skill, the same way `atlas/` was written directly by the skill this one replaces. Do not reference a CLI subcommand that doesn't exist. Run the 8 steps below, in order (design §6.4). ## Step 1 — Invocation The user hands you a document (path, pasted text, or upload) and, optionally, names its type by the `id` of an `archive/types/<id>.yaml` file. ## Step 2 — Classify Read every `archive/types/*.yaml` file. Each parses to this exact shape (`src/archive/formats.ts` `ArchiveTypeDef` — use these field names, no others): ```yaml id: client-spec # type id == filename stem; becomes metadata.yaml `kind` label: Client specification classification: extensions: [.pdf,