Skip to content

feat: universal freview skill — daily/full modes, config, JSON and specialists - #1

Merged
raiseCatError merged 82 commits into
mainfrom
feat/universal-agent-skill
Oct 16, 2026
Merged

raiseCatError merged 82 commits into
mainfrom
feat/universal-agent-skill

Conversation

@raiseCatError

@raiseCatError raiseCatError commented Oct 16, 2026 •

Copy link
Copy Markdown
Owner

freview v0.1 — universal evidence-first audit skill

Updated modes:

  • freview / freview standard: daily metadata-only health review, no implementation-source review and no deep code review; never modifies code.
  • freview full: complete applicable delivery/hygiene audit including source-level implementation tracing and deep code-quality review. Prefer an already-installed trusted codebase-review skill, otherwise native agent review. No automatic installs, edits or code uploads.
  • quick, only, excluding, --with security-scan, --base for narrower scope. Security scanners never execute without approval.

Project tailoring: validated, declarative .freview.json (versioned JSON Schema), optional development/release branches, limits, specialist preferences and human QA notes. No hooks, arbitrary commands or automatic config creation.

Reports: machine-readable v1 JSON Schema plus optional no-dependency CLI review.py (Markdown/JSON); distinguish local-snapshot from actual agent partial-audit/full-audit. Read-only report_tools.py validates and compares stable finding IDs across compatible daily reports. CLI --output requires an explicit new file path; no default writes.

Specialists: a curated catalog of codebase-review, OpenSSF Scorecard, OSV-Scanner, dependency-review, CodeQL, Semgrep, Gitleaks, zizmor, markdownlint-cli2 and lychee. Discovery hints only, never automatic installation/execution. Repolinter is archived and not a maintained default.

Automation: a separate opt-in GitHub Actions scheduled local structural snapshot template. This does not start paid AI review or install itself in consumer repos. A scheduled agent-assisted daily report requires a separately authorized scheduler/agent connection.

Packaging: the full Agent Skill is self-contained under skills/freview/ including its scripts, linked refs, schemas, specialist registry and automation template. Repository-level copies exist for developer convenience and are checked for identical bytes.

Safety: independent of NMSh; no deletes, branch pruning, fetch/checkout, CI setup on other repos, merges, releases or issue edits. Source and issue content are untrusted. Active user storage policy remains in force.

Automated checks: Python standard-library fixtures for clean/dirty/multi-worktree Git, config validation and precedence, daily vs full defaults, JSON report and comparisons, plus Agent Skill packaging checks on Ubuntu/macOS. Check latest PR head CI before merging. Real Claude/Codex/OpenCode/Gemini installer discovery still requires a hands-on smoke test.

NMSh draft PR #400 contains an earlier scoped prototype; reconcile it as a consumer after upstream acceptance. Do not automatically merge or publish v0.1.

@raiseCatError raiseCatError changed the title feat: portable freview agent skill with modes, hygiene audit and optional specialists feat: universal freview skill — daily/full modes, config, JSON and specialists Oct 16, 2026
@raiseCatError
raiseCatError marked this pull request as ready for review October 16, 2026 18:46
@raiseCatError
raiseCatError merged commit 4d2ba18 into main Oct 16, 2026
2 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant