Releases: fullstackcrew-alpha/privacy-mask
Releases · fullstackcrew-alpha/privacy-mask
v0.3.5 — Fix ClawHub security flag
- Exclude SUBMISSIONS.md from ClawHub publish (contained example API key strings that triggered security scanner)
- Add references/ and SUBMISSIONS.md to .clawhubignore
v0.3.4 — Improve discoverability & fix ClawHub security flag
Changes
- Declare
UserPromptSubmithook in SKILL.md metadata (fixes ClawHub "suspicious patterns" flag) - Expand PyPI keywords from 5 to 19 terms for better search visibility
- Add PyPI classifiers (Image Recognition, Console)
- Enhance SKILL.md with vector-search synonyms (anonymize, GDPR, DLP, compliance)
- Add numbered steps, conditional logic, constraints, and anti-patterns sections
- SkillsMP quality score: C (62/100) → A+ (90/100)
- Add
references/directory - Add 5 new GitHub repo topics (redaction, tesseract, gdpr, secret-detection, pii)
- Sync all version locations to 0.3.4
v0.3.2 — Further reduce NER false positives
Changes
- Raise NER confidence threshold from 0.6 to 0.7
- Add min_word_count filter for person names: single-word matches (browser tabs, usernames, button labels) are now skipped. Only multi-word names like "Zhang San" or "John Smith" are flagged.
- Add Troubleshooting docs for
UserPromptSubmit hook errortimeout issue (recommend 60s timeout for NER engine)
Upgrade
pip install --upgrade privacy-maskv0.3.1 — Reduce false positives
Changes
- Disable 8 high false-positive regex rules by default: IP_ADDRESS, UUID, MAC_ADDRESS, IPV6, QQ_NUMBER, URL_AUTH_TOKEN, MRZ_LINE1, MRZ_LINE2. Rules remain in config.json with
enabled: false— re-enable via custom config if needed. - Remove
organization nameandlicense plate numberfrom default NER entity types — these triggered on Apple/Google/app names in UI screenshots. - Raise NER confidence threshold from 0.5 to 0.6 to further reduce low-confidence false positives.
Upgrade
pip install --upgrade privacy-maskv0.3.0 — NER Detection Engine + Community Infrastructure
What's New
NER Detection Engine (GLiNER)
- Zero-shot Named Entity Recognition via GLiNER as an alternative detection engine
- Detects person names, street addresses, organizations, dates of birth, medical conditions, and more — without regex
- Default engine with automatic fallback to regex when
glineris not installed - Install with:
pip install privacy-mask[ner] - Switch engines via CLI:
--detection-engine regex|ner - Configure in
config.json:"detection": { "engine": "ner" }
Community Infrastructure
- Issue templates: bug report, feature request, new detection rule
- Pull request template with checklist
- Code of Conduct (Contributor Covenant)
Other
- Updated architecture diagram reflecting dual-engine design
- Updated README with detection engine documentation
- Version sync: 0.3.0 across pyproject.toml, SKILL.md, marketplace.json
Install
# Regex only
pip install privacy-mask
# With NER engine (recommended)
pip install privacy-mask[ner]v0.2.0
What's New
Fixed
- Fix MRZ_LINE2 false positives — require chevron separators
- Fix SWIFT_BIC false positives — add ISO 3166 country code constraint
Added
privacy-mask on/off/statustoggle commands- Claude Code plugin structure
- GitHub Actions CI + auto-publish to PyPI
- 208+ tests, CONTRIBUTING.md, CHANGELOG.md
- Animated GIF demo and SVG architecture diagram
Install: pip install privacy-mask==0.2.0