v1.2.0 — Inbox lifecycle, bundle aliases, knowledge-base commands

Fixes six defects found in the first real end-to-end run of the inbox → OKF knowledge-base pipeline.

P0 fixes (blocked documented workflows):

• The inbox could never be cleared — five files contradicted the sixth's "keep items here only until classified" intent, so an agent following the contract correctly refused to remove anything. Now a scoped clear authority: an inbox copy may be removed once its content is verifiably filed at the destination named in the approved triage table, never before that destination copy exists.
• Every bundle wikilink broke on the first ingest — bundle indexes are all named index.md, so [[bundle-slug]] could never resolve by filename, and bundle-template.md shipped no aliases: field. Fixed, with setup/verify.py now failing mechanically on a bundle index missing its alias.

P1 fixes:

• Documented that inbox material bound for the knowledge base goes to 04-knowledge/raw/<type>/ and is then ingested, not just copied in.
• Added /ingest, /query-knowledge, /lint-knowledge commands and routing rows so the knowledge base's operating model is actually invokable.
• The wiki (the shareable layer) now has an automated anonymity gate: scripts/privacy_check.py scans 04-knowledge/WorkOSWiki and 04-knowledge/schema.

P2 fix:

• Re-derived all 14 built 04-knowledge/ files from setup/templates/, restoring content that had drifted out (including the aliases rule that would have prevented the bundle-link bug). Added tests/drift/test_template_drift.py so this class of drift fails the test suite going forward.

Verified end to end on a clean copy: a document dropped into 00-inbox/ reaches a bundle as a linked OKF page with a provenance page, updated indexes, and a log entry — ending with an empty inbox, no manual file edits, and all checks green.

Full details: 4cb6029

v1.1.0 — Your Private AI Work OS

Reframes the identity/privacy/anonymity rule so it reads as protection for the shareable instruction layer, not daily governance over your own data.

Highlights

• Renamed the reader-facing product to Your Private AI Work OS (README, CLAUDE.md, setup titles).
• Repositioned the anonymity rule below the Session protocol as "Keeping the shareable layer reusable" — it protects the shareable layer, not your 01-context/ data; you only need it when editing shared instruction files or contributing upstream.
• Optional, purpose-scoped tooling: /privacy-check and /audit-workos are labeled pre-share / contribution checks, not daily hygiene.
• Neutral routing language: the capture-and-classify "anonymity gate" is now "personal data goes in the personal zone."
• First-timer clarity in README, 01-context/README.md, and setup: real data in 01-context/ is expected, not a leak.
• Versioning: added a VERSION file and README version badge.

The privacy machinery (privacy_check.py, tests, protected-terms) is unchanged and still functional. No breaking structural changes.

v1.0.0 — Anonymous AI Work OS

First public release of WorkOS — a portable, file-based AI Work OS template for professional knowledge workers.

What's inside

• Anonymous instruction layer kept separate from your private 01-context/ (only *.example.* templates ship; real context is git-ignored).
• 4 agents — research-analyst, knowledge-librarian, planner, quality-reviewer.
• 6 skills, 7 commands, 4 workflows with explicit approval points and dry-run defaults.
• A Session protocol in CLAUDE.md so any agent orients itself and picks the right skill.
• Privacy & structure tooling — privacy_check.py, validate_structure.py, and a runnable test suite.
• Optional Obsidian/OKF knowledge wiki scaffold under 04-knowledge/.
• Setup guides for building the OS, adding the wiki, and wiring it to any Claude surface.

Getting started

Click “Use this template”, then follow setup/SETUP.md. Verify with python3 tests/run_tests.py.

License

CC BY-SA 4.0 © Steff Vanhaverbeke — credit the original author and share adaptations under the same license.