jeremylongshore

hyperflow-audit

Hyperflow code review. Use when the user wants the current diff, a commit, branch, or PR reviewed — verbs like audit, review, "check for issues", "security check", "code review". Multi-level review (L1 quick → L5 exhaustive), writes findings to .hyperflow/audits/, then a fix-gate.

jeremylongshore 2,679 392 Updated 1mo ago
GitHub

Install

npx skillscat add jeremylongshore/claude-code-plugins-plus-skills/hyperflow-audit

Install via the SkillsCat registry.

SKILL.md

hyperflow-audit — review phase (Antigravity single-agent)

Multi-level review over a target (default: git diff HEAD + staged). Follow the hyperflow doctrine. Security scan is mandatory at L3+.

Levels

L Checks
1 syntax, obvious bugs, formatting
2 L1 + spec compliance, naming, edge cases
3 L2 + cross-file consistency, integration risks, security (secrets, injection, path traversal, XSS, missing validation)
4 L3 + architecture, scalability, accessibility
5 L4 + adversarial probing, perf profiling, alternatives

Default to L2; elevate to L3 when the diff touches auth, data, money, or external input.

Steps

  1. Resolve scope (target arg or current diff). Read the changed files + their immediate dependencies.
  2. Review at the chosen level. Grade each finding [Critical] / [Important] / [Suggestion] / [Praise] with file:line + a concrete fix.
  3. Write the full report to .hyperflow/audits/<YYYY-MM-DD-HHmm>-<scope>.md (status table → TL;DR → findings → security-scan table). Print a one-line summary pointing at the file.
  4. Fix gate via AskUserQuestion (only when Critical/Important exist): Fix all (Recommended) / Critical+Important / Critical only / No. On a fix choice, route the findings into hyperflow-planhyperflow-dispatch. On SECURITY_VIOLATION, skip the gate and surface immediately.

Rules

  • Findings live in the file, not chat — chat shows only the summary box.
  • A clean run (no Critical/Important) prints Audit clean and still writes the file for history.