failure-analysis-package

Analyzes git-cg commit-message gold misses using METHOD-compliant forensic case documentation.

2|Updated Jun 7, 2026
One-click install
npx skills add https://github.com/Thomo1318/gitCommitGenerator --skill failure-analysis-package-thomo1318
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: failure-analysis-package
Source: https://github.com/Thomo1318/gitCommitGenerator/tree/main/docs/quality/skill
Command: npx skills add https://github.com/Thomo1318/gitCommitGenerator --skill failure-analysis-package-thomo1318

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? When the git-cg commit generator produces a message that misses the gold standard, teams need a rigorous, repeatable way to reconstruct why the failure happened, classify it, and prevent recurrence. This Skill operationalizes the docs/quality failure-analysis package so every gold miss becomes a durable, indexed forensic case instead of a lost chat note. ## Core Features & Use Cases - Forensic Reconstruction: Walks the 14-stage commit pipeline (extract, rank, contract lock, render, accept) to identify the first stage that left the gold path, with Opik trace binding or explicit Opik-unbound labeling. - Regime & Taxonomy Classification: Assigns Regime A (controls fire but recovery worsens output) or Regime B (controls never fire and a wrong envelope is accepted), attaching F*/P* IDs only from the official FAILURE_TAXONOMY and PREVENTION_BACKLOG registries. - Case Promotion & Indexing: Lands durable cases under docs/quality/cases/, updates indexes and source maps, and supersedes GitHub issue comments as intake-only evidence. - Use Case: After a generated commit message passes Hybrid parsing but uses the wrong type and scope, use this Skill to file a full Regime B case with raw-to-gold SHAs, tree-preservation proof, and prevention control pointers. ## Quick Start Ask the agent to run a gold-miss failure analysis on the latest git-cg commit and file a case under docs/quality/cases following the METHOD protocol.

Frequently Asked Questions about failure-analysis-package

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

FAQPage Schema
How do I analyze a git-cg commit message gold miss?

Load docs/quality/METHOD.md, establish raw and gold commit identity with git commands, bind Opik traces or mark Opik-unbound, then walk the 14 pipeline stages to find where the message left the gold path. File the case using CASE_TEMPLATE.md under docs/quality/cases/.

What is the difference between Regime A and Regime B failures?

Regime A means controls fire but recovery makes output worse, such as guard thrash leading to skeleton fallback rewrites. Regime B means controls never fire, so a wrong type or scope envelope is cleanly accepted because gold_strict only scores shape.

When should I not use this failure analysis skill?

Do not use it for generic code review without a gold-miss case, product feature implementation, or inventing new presentation metrics inside the eval plan. It is strictly for commit-message gold-miss forensics governed by the docs/quality package.

Can I do a message-only commit rewrite without changing the tree?

Yes, but the tree OID must remain identical between raw and gold commits, verified with git rev-parse. Do not rely on --no-verify alone; use git commit-tree or the documented GIT_CG_SKIP_PREPARE flag per the message-only-rewrite process doc.

Why must Opik trace IDs never be fabricated in a case?

Fabricated spans destroy forensic integrity and mislead prevention work. If no trustworthy Opik bind exists, the case must explicitly state Opik-unbound with the reason, keeping evidence confidence honest.