Refined the musafety-tools status block to be easier to scan at a glance,
with TTY-enhanced separators/colors and a plain-text fallback for logs/CI.
The command list remains concise and aligned with the core help output.

Constraint: Keep output copy-paste friendly in non-interactive environments
Rejected: Full rich/verbose command docs in status output | too noisy for regular status checks
Confidence: high
Scope-risk: narrow
Reversibility: clean
Directive: Keep command summary concise in status; reserve detailed docs for README/help only
Tested: npm test (31/31 passing)
Not-tested: manual Windows terminal color rendering

Added a new README screenshot section showing the musafety status/log
presentation, using a terminal-style SVG that mirrors the current command
layout and recommendation footer.

Constraint: Keep README visuals lightweight and repository-local
Rejected: PNG screenshot artifact | larger binary and harder to diff/update
Confidence: high
Scope-risk: narrow
Reversibility: clean
Directive: Update this SVG when status log command wording changes materially
Tested: npm test (31/31 passing)
Not-tested: npm README rendering preview

This introduces managed template files for a local Codex musafety skill and
Claude /musafety command, then wires them into install/fix flows so setup
automatically provisions them and doctor repairs them if missing.

Constraint: Keep install/fix behavior idempotent and repo-local
Rejected: Writing skills into user home directories | harder to test and risky for shared environments
Confidence: high
Scope-risk: moderate
Reversibility: clean
Directive: If skill/command behavior changes, update templates and setup verification expectations together
Tested: node --test test/install.test.js test/metadata.test.js (31/31 passing)
Not-tested: manual Claude /musafety command execution in live Claude client

Bumped package version from 0.4.6 to 0.4.7 so the next publish can be
cut as a new npm release.

Constraint: Keep the release bump isolated to package manifest versioning
Rejected: Bundling unrelated pending workspace edits in this release bump | increases release risk
Confidence: high
Scope-risk: narrow
Reversibility: clean
Directive: Publish this version only after required CI and PR checks are green
Tested: npm pack --dry-run (musafety@0.4.7 tarball generated)
Not-tested: npm publish

CI jobs were failing before tests because setup-node npm cache and npm ci both require a lockfile. This repo's PR branch state currently does not guarantee a committed lockfile, so the workflow now installs dependencies with npm install --ignore-scripts and avoids cache mode that hard-fails on missing lockfiles.

Constraint: CI must stay green for pull_request merges even when package-lock.json is absent
Rejected: Commit package-lock.json as a required fix | branch currently carries unrelated local changes and lockfile enforcement was not an explicit repo policy
Rejected: Keep npm ci and add conditional lockfile logic | more workflow complexity than needed for this immediate failure mode
Confidence: high
Scope-risk: narrow
Reversibility: clean
Directive: If lockfile policy becomes mandatory, re-enable npm ci together with committed lockfile governance
Tested: node --test test/install.test.js
test/metadata.test.js
Tested: node --check bin/multiagent-safety.js
Tested: npm install --ignore-scripts
Not-tested: Full GitHub Actions rerun after push

Added two timestamped reports from the provided Scorecard snapshot:
a baseline score/check breakdown and a prioritized remediation plan for
raising the repository score.

Constraint: Report based on screenshot evidence provided in-session
Rejected: Live scorecard fetch from network | user supplied authoritative snapshot to document now
Confidence: medium
Scope-risk: narrow
Reversibility: clean
Directive: Refresh these reports after each major scorecard re-run to keep deltas current
Tested: Manual review of captured check scores against screenshot
Not-tested: automated scorecard ingestion pipeline