feat(okr): innboks-ingestion 1.7.0, docs + release (SC alle)

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

CLAUDE.md

@@ -1,4 +1,4 @@
-# OKR Offentlig Sektor v1.6.1
+# OKR Offentlig Sektor v1.7.0
 
 Expert OKR guidance for Norwegian public sector. Google/Doerr methodology adapted for 4-month tertial cycles.
 
@@ -16,6 +16,7 @@ Expert OKR guidance for Norwegian public sector. Google/Doerr methodology adapte
 | `/okr:gap` | Automatic gap analysis between tildelingsbrev and current OKR |
 | `/okr:analyse` | Cross-cycle analytics with Mermaid trend visualizations |
 | `/okr:oppsett` | Configure plugin: onboarding interview (full/mvp), cycle archival, profile update. Args: `full\|mvp\|arkiver\|oppdater\|vis` |
+| `/okr:innboks` | Ingest documents dropped in `.claude/okr/innboks/` into the OKF tree — deterministic convert/split/frontmatter/index, per-document security gate |
 | `/okr:export` | Export OKR deliverables (quality review, gap matrix, status report, retrospective) to print-ready PDF via weasyprint |
 | `/okr:freshen-references` | KB self-evaluator: score the 16 domain reference files against an anchored rubric + currency-poll public sources |
 | `/okr:help` | Full overview of all commands, agents, and recommended cycle workflow |
@@ -61,7 +62,7 @@ Cycle archival: `/okr:oppsett arkiver` — moves `syklus/` to `historikk/`, gene
 
 ## OKF Knowledge Layout
 
-Context files carry OKF-compatible frontmatter (Knowledge Catalog "Documents/kb Layout", aka "Metadata as Code" — not a formal "OKF v0.1" standard): a Title-Case `type` (`Organisasjonsprofil`, `Tildelingsbrev`, `Virksomhetsplan`, `Overordnede OKR`, `OKR`, `Retrospektiv`, `Status`) plus recommended `resource`/`title`/`description`/`tags`/`timestamp`. Only `type` is de-facto required. The flat frontmatter parser reads only single-line keys (`type`, profile/config fields); tree files may carry multi-line `tags` lists without breaking it.
+Context files carry OKF v0.1 Documents-style frontmatter (conforms to the OKF v0.1 minimal contract, canonized in `catalog/docs/okf-second-brain/spec.md`): a Title-Case `type` (`Organisasjonsprofil`, `Tildelingsbrev`, `Virksomhetsplan`, `Overordnede OKR`, `OKR`, `Retrospektiv`, `Status`) plus recommended `resource`/`title`/`description`/`tags`/`timestamp`. Only `type` is de-facto required. The flat frontmatter parser reads only single-line keys (`type`, profile/config fields); tree files may carry multi-line `tags` lists without breaking it. Files written by inbox ingestion additionally carry the provenance extension key `kilde: innboks`.
 
 **Two bundle roots** with distinct lifecycles: project `.claude/okr/` (cwd-bound cycle/work) and home `~/.claude/okr/org/` (reinstall-surviving org identity). Each root carries its own `index.md` (no frontmatter; `# H1` + `* [title](link) - description`) and an `okf_version` marker. `okf-index`/`okf-check` run per root; retrieval Globs both (project preferred, else home).
 
@@ -69,6 +70,7 @@ Context files carry OKF-compatible frontmatter (Knowledge Catalog "Documents/kb
 - `scripts/okf-index.mjs` — regenerate a root's per-level `index.md` (verbatim OKF index format).
 - `scripts/okf-check.mjs` — validate each concept file carries `type:` (exit 1 + count otherwise); report `okf_version` per root.
 - `scripts/compose-org-profile.mjs` — build the nested org-profile YAML with top-level OKF keys (`type`/`resource`/`timestamp`) before piping to `write-org-profile.mjs`.
+- `scripts/innboks-ingest.mjs` — inbox-ingestion orchestrator (`/okr:innboks`): walks `.claude/okr/innboks/`, converts (txt/md/docx/eml/pdf via `lib/innboks-convert.mjs`), heading-splits, stamps OKF frontmatter + `kilde: innboks`, routes by type, then gates each document with `okf-check --strict-ingest` — a failing document is discarded alone (staging + per-document rollback), survivors get relations and index entries. Non-destructive (originals stay in the inbox), idempotent by construction (timestamp = source mtime), path-confined (realpath). Binary formats need `npm install --ignore-scripts` (4 exact-pinned pure-JS deps); txt/md run dependency-free.
 
 Retrieval is on-demand via the `okr-second-brain-search` skill. The UserPromptSubmit hook no longer pre-injects the full context-file enumeration — it emits the core profile plus one resolved pointer to `index.md` (< 512 B, independent of file count).
 
@@ -92,6 +94,7 @@ Retrieval is on-demand via the `okr-second-brain-search` skill. The UserPromptSu
 /okr:analyse ──→ trendanalytiker
 /okr:oppsett ──→ (inline wizard: full/mvp/arkiver/oppdater/vis)
 /okr:oppsett arkiver ──→ cycle archival + retrospektiv-generering
+/okr:innboks ──→ scripts/innboks-ingest.mjs (convert → split → frontmatter → per-doc gate → relations → index)
 /okr:export ──→ scripts/export-pdf.py (weasyprint, documented prerequisite)
 /okr:freshen-references ──→ (inline KB-evaluator + currency-polling via WebSearch/Task)
 /okr:help ──→ (inline command/agent/workflow overview)