triage

Diagnose bugs and test failures into verified root causes with evidence-backed fix proposals.

Updated Jun 15, 2026
One-click install
npx skills add https://github.com/nsollazzo/ship-kit --skill triage-nsollazzo
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: triage
Source: https://github.com/nsollazzo/ship-kit/tree/main/.agents/skills/triage
Command: npx skills add https://github.com/nsollazzo/ship-kit --skill triage-nsollazzo

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve? When a bug, test failure, or unexpected behavior is reported, jumping straight to a fix often treats the symptom instead of the cause. This Skill enforces a disciplined, read-only investigation that traces the bad value back to its origin and verifies the root cause with concrete evidence before any fix is proposed. ## Core Features & Use Cases - Parallel investigation angles: Fans out sub-agents (or sequential passes) across reproduction, recent changes via git history, backward root-cause tracing, and blast-radius analysis. - Adversarial verification: Spawns an independent sub-agent whose job is to refute the leading hypothesis, so only evidence-backed root causes survive. - Source-level fix proposals: Recommends fixes at the origin of the bug, with optional defense-in-depth hardening guidance, and hands off to the fix skill for implementation. - Use Case: A CI test suite starts failing after a dependency update. Invoke triage with the failing test output; it reproduces the failure, traces the bad value through the call chain, verifies the root cause against git diffs, and reports the exact file:line evidence plus a recommended fix. ## Quick Start Ask the agent to triage the failing test or bug report by pasting the error message or issue description and requesting a root-cause diagnosis.

Frequently Asked Questions about triage

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I find the root cause of a bug instead of fixing the symptom?▼

Trace the bad value backward up the call chain to where it first appears, not where the error surfaces. This Skill fans out investigation angles including reproduction, git history, and root-cause tracing, then verifies the leading hypothesis adversarially before proposing a fix.

How to debug a failing test with AI sub-agents?▼

Provide the failing test output or error message as the argument. The Skill spawns parallel sub-agents across angles like reproduction, recent changes, and blast-radius, then reconciles findings into one verified root cause with file:line evidence.

Does the triage skill modify my code during investigation?▼

No, triage is strictly read-only. It produces a diagnosis with a recommended fix but never edits code; applying the fix is handed off to the separate fix skill, which handles branching, implementation, and shipping.

What if every proposed fix just moves the problem elsewhere?▼

That pattern signals a design problem, not a patching problem. The Skill instructs the agent to stop and surface it as an architecture question to the requester rather than recommending a fourth workaround fix.

When should I not use a full multi-agent triage investigation?▼

Skip the full fan-out for trivial issues like a one-line typo, where directly reading the code is faster. The investigation scales to the issue: sub-agents are warranted for cross-component failures, not obvious defects.