elite_build_understanding

Validates builds through stage-gated design, plan, build, test, and ship checks.

1|Updated Jun 17, 2026
One-click install
npx skills add https://github.com/Tcuzzo/HydraAgent_public --skill elite-build-understanding-tcuzzo
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: elite_build_understanding
Source: https://github.com/Tcuzzo/HydraAgent_public/tree/main/skills/elite_build_understanding
Command: npx skills add https://github.com/Tcuzzo/HydraAgent_public --skill elite-build-understanding-tcuzzo

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Builds often drift from the original intent, pass shallow tests, or ship with fabricated claims about what was implemented. This Skill provides a stage-gated validation discipline that interrogates a build from intent through delivery, using live understanding organs that return evidence rather than rubber-stamping work. ## Core Features & Use Cases - Stage-Gated Validation: Runs Design, Plan, Build, Test, and Ship gates, each bound to live validator calls that return scores, verdicts, failures, and recovery actions. - Intent Deduction: Translates compressed human language (metaphor, slang, dialect) into a stated directive before scoring, avoiding literalism and caricature failures. - Drain-Refute Honesty: Runs refute_fabricated_claim on any claim referencing symbols, quotes, files, or test results, driving fabricated claims to a failing score. - Use Case: Before shipping a feature, run the build through the stage gates to confirm the spec matches intent, tests exercise real behavior without mocking the validator under proof, and failures are repaired before landing. ## Quick Start Validate my current build against the original intent using the design, plan, build, test, and ship stage gates.

Frequently Asked Questions about elite_build_understanding

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

FAQPage Schema
How do I validate a build against the original intent?▼

Run the build through the five stage gates: Design, Plan, Build, Test, and Ship. Each gate calls a live validator that returns evidence such as scores, verdicts, and named failures, and you revise the work from that evidence rather than inventing gate results.

How to detect fabricated claims in AI-generated code summaries?▼

Run refute_fabricated_claim on any claim referencing a symbol, quote, file, or test result. Every referenced item is resolved against the installed artifact, and any unresolved item refutes the claim, drives the score to the floor, and sets passed to false.

What is the drain-refute honesty check in build validation?▼

Drain-refute honesty requires that a validator can refute false claims, not merely confirm them. An organ that can only agree is an echo; the refute path against real artifacts is what makes the validation honest.

When should I mock dependencies in build validation tests?▼

Mock only the outermost unstable leaf, such as replacing a model call with a canned verdict. Never mock the validator whose scoring and pass/fail logic is the proof, since that produces a green-but-wrong test that only confirms its own canned answer.

Does a high confidence score mean the build is ready to ship?▼

No. Confidence alone never establishes readiness. Tests and the real surface proof decide whether the capability is done, and every named failure must be repaired with the returned recovery actions before rerunning the validator.