systematic-bugfinder

Diagnose and fix software bugs across five layers from symptom to systemic safeguard.

7|Updated May 13, 2026
One-click install
npx skills add https://github.com/medialab/spinosa --skill systematic-bugfinder-medialab
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: systematic-bugfinder
Source: https://github.com/medialab/spinosa/tree/main/.agents/skills/systematic-bugfinder
Command: npx skills add https://github.com/medialab/spinosa --skill systematic-bugfinder-medialab

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve? Debugging often stops at the first passing test, leaving the root cause, neighboring defects, and recurring bug families unaddressed. This Skill enforces a disciplined five-layer investigation so every fix is evidence-based, regression-tested, and hardened against recurrence. ## Core Features & Use Cases - Five-Layer Bug Sweep: Systematically works from the reported symptom (L1) through direct cause (L2), neighboring paths (L3), recurring patterns (L4), and systemic safeguards (L5). - Finding Ledger & Audit Dispositions: Tracks every candidate bug with evidence, confidence, and a disposition such as fixed, stale, intentional, duplicate, or test gap. - Race-Condition & Regression Discipline: Provides a race-condition test matrix, hypothesis queue, and stopping rules to prevent speculative patches and circular debugging. - Use Case: When a user reports that switching workspaces shows stale data, the Skill reproduces the failure, traces the async identity bug, checks sibling code paths for the same pattern, adds a regression test, and reports a disposition for every finding. ## Quick Start Ask the agent to systematically investigate and fix the reported bug, including any related instances of the same root cause.

Frequently Asked Questions about systematic-bugfinder

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

FAQPage Schema
How do I systematically debug a software bug instead of guessing?▼

Start by reproducing the exact symptom, then trace one causal chain from entry point to the first wrong state. Test one hypothesis at a time with a discriminating check, fix the shared primitive rather than individual callers, and add a regression test that fails before and passes after the fix.

How to find all bugs related to the same root cause?▼

After confirming the root cause, search by mechanism rather than symptom: inspect every caller of the changed primitive, sibling branches, retry and cancellation paths, identity switches, and persisted state. Fix only confirmed neighboring defects and record dispositions for the rest.

What is the difference between diagnose, fix, and audit modes?▼

Diagnose identifies and explains the cause without code changes. Fix diagnoses, implements, and verifies the reported bug. Audit classifies a supplied finding list as confirmed, stale, intentional, hypothetical, test gap, or duplicate, then fixes only confirmed defects.

How do I test async race-condition fixes?▼

Test cancellation or identity switches before and during awaits, completion after invalidation, a new run starting while the old one unwinds, and concurrent commits to the same destination. Use generation-scoped tokens and serialize writes to prevent overlapping commits.

When should I stop widening a bug investigation?▼

Stop when the original symptom is reproduced and resolved, one causal chain explains the failure, all consumers of the changed primitive are inspected, two consecutive pattern searches find nothing new, and all regression and maintained-suite checks pass.