Changelog

0.1.1 · Less noise

November 7, 2026 · Release notes on GitHub

Fewer false alarms on real projects. Running on a folder of seven projects, findings dropped from 153 to 68 with no real problem lost.

Engine and rules

• Skips build output, caches and minified files
• New ignore list in lintsure.config.yaml to skip paths such as tests or fixtures
• Personal data in logs: words inside a message no longer count, values put into it still do
• TLS checks: verify=False is only flagged on HTTP client calls

0.1.0 · First public release

November 7, 2026 · Release notes on GitHub

The engine, CLI, MCP server and GitHub Action are open source. 16 draft rules across the EU AI Act, GDPR, SOC 2 and ISO/IEC 27001, with a human review step built into every result.

Engine and rules

• Core engine, CLI and MCP server, with tests that run the same input twice and expect the same result
• 8 GDPR and SOC 2 rules, each with a planted example and a near miss that must not be flagged
• 5 EU AI Act rules, and rules that can read a whole file
• Plain-English knowledge of each framework, used to suggest frameworks and list work outside code
• Choose frameworks with lintsure.config.yaml

Human in the loop

• Every report and every MCP reply carries the disclaimer, the work outside code and the next steps for a person
• The MCP server tells coding agents not to give verdicts, not to decide, and to ask before any change
• Tests fail if Lintsure's own wording claims you are compliant, certified or ready

Reports

• Multi-page assessment report with findings, possible fixes, an outside-code checklist and a human sign-off page (lintsure scan --report)
• Clearer CLI summary line

GitHub Action

• First version of the action
• Builds the engine from source, so it works without npm; self-test that a planted problem fails the job

Not yet

• Auditor-reviewed rule mappings
• The team platform