All notable changes to @cobusgreyling/loop-audit are documented here.
- Harness Runtime scoring signals for the LE → harness-foundry funnel:
.foundry/stack.yaml,stack.lock, sessions/traces, outerloop emit path, host integrate hints
- Loop Ready 80+ recommendation / human report CTA →
loop-init --with-foundry --suggestcopy commands for Foundry scaffold + firstfoundry run
- Governance scoring signals: least-privilege tool scope, stall / no-progress detection, human-escalation path
allowed-toolsfrontmatter inSKILL.mddetected as a least-privilege signal--suggestrecommendations when governance signals are missing
- Contributor quickstart CTA after CLI runs (skipped for
--json,--md, and--badge)
- Detect verifier agents in
opencode.jsonandopencode.json.example(maker/checker split for Opencode loops) --suggestcopy commands for Opencode (loop-init --tool opencode)
loop-constraints.mdandloop-constraintsskill detection in readiness scoring (+6 points when both present)- Recommendations when constraints file or skill is missing
- Updated package description and keywords for better discoverability on npm / npx (emphasizes "loop engineering", coding agents, and concrete usage examples).
- Unit tests for scoring logic (
test/auditor.test.ts) --suggestnow mentionsloop-initscaffold CLI- Registry and starter coverage in audit recommendations
- CI gates on test suite before publish
--suggestcopy-from-template commands for Grok, Claude Code, and Codex- Expanded signals: MCP, worktree evidence,
patterns/registry.yaml - L3 scoring threshold with verifier + state requirements
--mdmarkdown report format- Safety doc and GitHub workflow detection
- Initial Loop Readiness Score CLI (L0–L3)
--jsonoutput for CI integration