Agents
Give agents evidence, not intuition.
Run one scan, read completeness, then fix the top-ranked issue. Agents must recover incomplete scanners before they treat a score as authoritative.
npx @neuralaxis/vibedoctor scan --report agent-json That JSON is the simplest agent input: completeness, topIssues, rootCauseGroups, recoveryActions.
One-off handoff
Then write the plan file.
npx @neuralaxis/vibedoctor agent-plan --format markdown Default artifacts under .vibedoctor/:
.vibedoctor/
├── report.json # automation and agents
├── report.html # human review
└── agent-plan.md # ordered repair workProtocol
What an agent is supposed to do.
- Read
completeness. Recover tools if the scan is PARTIAL or INVALID. - Fix the top-ranked issue in
topIssues. Use location, repair, and verification fields. - Preserve unrelated user changes.
- Verify with the listed tests and
vibedoctor scan --changed --report agent-json. - Continue down
recommendedOrder. Never claim success from a PARTIAL or INVALID scan.
Agent pack
Repository-scoped instructions.
npx @neuralaxis/vibedoctor agent init --targets all
Targets: codex, copilot, claude, cursor. This can generate AGENTS.md,
.agents/skills/, target-specific skills for Claude and GitHub,
Cursor rules and MCP config, and .vibedoctor/agent-policy.yml.
Check the wiring with npx @neuralaxis/vibedoctor agent doctor --targets all.
Installable plugin bundles for codex and claude:
npx @neuralaxis/vibedoctor agent plugin --targets all --force Fix order
The pack is opinionated on purpose.
| # | Fix first |
|---|---|
| 1 | Security blockers |
| 2 | Failing tests |
| 3 | Type errors |
| 4 | Dependency vulnerabilities |
| 5 | Privacy Review findings with regulated identifiers or sensitive attributes |
| 6 | High-confidence dead code |
| 7 | Leftovers and legacy baggage |
| 8 | Refactor-readiness items |
| 9 | Efficiency suggestions |
Do not delete dead code unless it is high confidence and independently verified. Do not send raw personal data to an AI service unless config and the user allow it.
MCP
Structured operations over stdio.
npx @neuralaxis/vibedoctor mcp | Tool | Does |
|---|---|
| vibedoctor_scan_changed | Run VibeDoctor on changed files and return normalized findings. |
| vibedoctor_scan_full | Run a full repository scan. |
| vibedoctor_get_report | Retrieve the current normalized report. |
| vibedoctor_explain_finding | Explain a single finding by ID. |
| vibedoctor_fix_safe | Apply supported safe tool fixes. |
| vibedoctor_get_agent_plan | Return the current repair plan in JSON or markdown. |
| vibedoctor_verify | Re-run a changed-scope scan after edits. |
| vibedoctor_dpdp_scan | Run a DPDP technical-readiness scan. |
| vibedoctor_dpdp_scan_changed | Collect DPDP evidence from changed files. |
| vibedoctor_dpdp_get_data_map | Return the personal-data map. |
| vibedoctor_dpdp_get_control_matrix | Return the control matrix. |
| vibedoctor_dpdp_get_review_queue | Return questions that require human review. |
| vibedoctor_dpdp_explain_control | Explain a control or finding. |
| vibedoctor_dpdp_get_evidence | Return the evidence ledger. |
| vibedoctor_dpdp_verify | Verify DPDP technical readiness after changes. |
| vibedoctor_dpdp_generate_handoff | Generate a DPDP agent handoff. |