privacy-abstractionlisted

# privacy-abstraction ## Mission Rewrite content so it preserves meaning while removing identifying or sensitive details. Used in `abstract` privacy mode. ## Source files (per ZEREF_OS §4) - `PRIVACY.md` (root) — modes - `REDACT.md` (root) — concrete classes (the authoritative list of what to strip) ## Rules (driven by REDACT.md `classes:` block) For each `enabled: true` class in `REDACT.md`, apply the replacement strategy: | Class | Strategy | |---|---| | credentials | full removal → `<REDACTED:credential>` | | pii | hash-based pseudonym → `<user:a3f9>` (consistent across writes within session) | | internal_paths | abstract to repo-relative → `/Users/x/proj/foo` → `<repo>/foo` | | client_data | full removal or pseudonym (project-configured) | | financial | bucket — `$1,234,567` → `<order:$1M-10M>` | | proprietary_code | full removal → `<REDACTED:internal>` | Custom classes appended to `REDACT.md` route through `skills/privacy-abstraction/SKILL.md`'s regex/heuristic library by `pattern` name. ## Operations ### REWRITE (called by privacy-guardian) 1. Receive payload + caller mode (`abstract` is the trigger) 2. Read enabled classes from `REDACT.md` 3. Apply replacements per strategy 4. Return cleaned payload + transformation log (what was replaced, which class matched) 5. The transformation log is itself stored in `memory/patterns/PATTERNS.jsonl` so the user can audit what was abstracted ### CONNECTOR_REWRITE (called by privacy-guardian for outbound transmission) 1.