vibeguard

name	vibeguard
description	Use VibeGuard anti-hallucination rules, hooks, guards, and verification practices for task startup checks, code review, risk assessment, and weekly review.

VibeGuard

VibeGuard prevents common AI-assisted development failures through compact rules, executable guards, hooks, focused tests, and evidence-based review.

Canonical sources:

• README.md — product entry and core/workflow boundary
• docs/rule-reference.md — public rule and guard summary
• schemas/install-modules.json — install/runtime contract
• rules/claude-rules/, hooks/, guards/, and vibeguard-runtime/ — implementation

Historical specs and plans are context, not an automatic work queue.

When to Activate

Use this skill when the user asks for VibeGuard, anti-hallucination checks, task startup constraints, guard rules, risk scoring, code review, or weekly review.

For ordinary implementation work:

• execute clear, bounded tasks directly;
• plan only major architecture or explicitly requested planning;
• keep normal specs to two files and about 300 lines total;
• avoid delegation unless the user asks or ownership is genuinely independent;
• use focused tests during iteration and broader checks before submission.

Red Flags

• A new rule has no executable guard, hook, test, or evaluation path.
• A clear small task is blocked on a process packet, schema, receipt, or runtime snapshot.
• Multiple writable sessions or agents operate on the same repository.
• A PR continues reviewing after Findings: 0 and PASS, or exceeds two initiated review rounds.
• A completion claim lacks fresh verification from the current session.

Checklist

• Confirm goal, context, constraints, and done-when.
• Search for existing rules, hooks, workflows, skills, and tests.
• Make the smallest requested change.
• Run the focused verification command for the changed behavior.
• Stop review immediately on zero findings and pass.
• Preserve unrelated worktree changes.

Seven-Layer Summary

Layer	Purpose
L1	Search before creating
L2	Naming and boundary conversion
L3	Hooks and fail-visible interception
L4	Architecture and code-quality guards
L5	Small delivery workflows
L6	Compact injected behavior rules
L7	Human review and trend feedback

References

• references/task-contract.yaml — startup checklist
• references/review-template.md — weekly review template
• references/scoring-matrix.md — risk/impact scoring
• workflows/references/delivery-base.md — delivery and review convergence rules

Guardrails

• Do not add prose-only enforcement.
• Do not invent APIs, fields, or success evidence.
• Do not swallow user-visible errors.
• Do not turn roadmap or historical documents into work without current user intent.
• Do not build a validator for another spec validator.