v0.4.1

Changelog

• e66e783: Release v0.4.1 (@0x1Adi)

v0.4.0

Changelog

• 7c50657: Refuse operator-only ai keys from discovered configs; add command provider; measure a local model as the judge (REPORT §15) (@0x1Adi)
• 3c9574f: Release v0.4.0 (@0x1Adi)
• 608dd44: add MCP install docs and marketplace icon, stop tracking session scratch (@0x1Adi)
• e25fc20: do not auto-approve the MCP server: scan_file reads any path (@0x1Adi)

v0.3.8

Changelog

• f00cd73: v0.3.7: MCP asks the agent when no provider is configured (@0x1Adi)
• 456583a: v0.3.8: ship an MCP Bundle with every release (@0x1Adi)

v0.3.6

Changelog

• e9db430: v0.3.6: MCP server matches the CLI, survives bad frames, negotiates versions (@0x1Adi)

v0.3.5

Changelog

• 38b7b4c: v0.3.5: pre-commit hooks ship off by default (@0x1Adi)

v0.3.4

Changelog

• 12f8834: v0.3.4: pre-commit framework hooks (@0x1Adi)

v0.3.3

Changelog

• 91c4d20: Add privacy policy (@0x1Adi)
• bb9d804: Plugin 0.3.2: load on current Claude Code (@0x1Adi)
• 8f3e161: Plugin 0.3.3: register the MCP server from .mcp.json (@0x1Adi)
• 0b4a082: Plugin 0.3.4: do not start the MCP server (@0x1Adi)
• 33e3f8a: Plugin 0.3.5: say plainly when klarion is not installed (@0x1Adi)
• 75b62fc: v0.3.3: hook never approves calls, scans commits, shows fingerprints (@0x1Adi)

v0.3.2

Changelog

• 1dae4d9: Action: pin the scanner to the release, list findings, survive rejected uploads (@0x1Adi)
• f21cbdc: ci: run the AI self-scan only on manual dispatch (@0x1Adi)
• eba094c: detect: clear gosec G101 and G115 (@0x1Adi)

v0.3.1

Changelog

• 1e7a643: Fixes creddata (@0x1Adi)
• ec0f18c: v0.3.1: test path rule, verdict cache fix, hook fixes, plugin marketplace, CredData results (@0x1Adi)

v0.3.0

Changelog

• ac57b65: benchmark: measure the AI configuration and record what it found (@0x1Adi)
• 00ee28b: benchmark: measure the AI configuration and record what it found (@0x1Adi)
• de8a298: detect: find format-less credentials without entropy; decode and rescan; history over all refs (@0x1Adi)
• 2371bfa: feat: forward reasoning_effort so a reasoning model does not bill its scratchpad (@0x1Adi)
• 6ed7723: fix: apply the request timeout per attempt so a retry-after wait cannot exhaust it (@0x1Adi)
• dde9fca: fix: honour fractional Retry-After so rate limits pause the scan instead of killing it (@0x1Adi)
• 9847d34: fix: ignore base64-encoded binaries by default (@0x1Adi)
• 3f2546d: fix: persist verdicts during a scan so an interrupted run resumes (@0x1Adi)
• 72f7f09: fix: print the model output that failed to parse (@0x1Adi)
• ab0aac8: fix: read the reasoning channel when content is empty (@0x1Adi)
• a61ec76: fix: survive provider rate limits, slow responses and unparseable batches (@0x1Adi)
• 36841b7: fix: survive rate limits and unparseable batches; show adjudication progress (@0x1Adi)
• ef4b52f: fix: survive rate limits and unparseable batches; show adjudication progress (@0x1Adi)
• bf5d7c2: verify: print adjudication progress from the start; harness streams it (@0x1Adi)