Hraness Engineering

Get evidence before you change the repository.

Hraness Engineering is 11 bounded Agent Skills for understanding, verifying, testing, documenting, measuring, and safely refactoring one repository concern. Each workflow owns a distinct result, so the agent can choose a method without blurring analysis, implementation, and delivery.

First proof

Ask:

Map this repository’s user-visible features and proof paths.

repository-feature-map returns a navigable map of domains, entry points, runtime flows, data ownership, and validation surfaces. It does not turn the request into an implementation project or invent historical rationale.

How it works

task → bounded engineering owner → authoritative repository sources → focused proof

Skill	Result
`repository-feature-map`	A navigable map of repository domains, entry points, flows, and validation surfaces
`semantic-code-analysis`	An evidence-backed explanation of how code behaves now
`reconstruct-rationale`	A dated, confidence-calibrated reconstruction of why code reached its current shape
`assess-change-impact`	A blast-radius and compatibility-risk register for a defined change
`compare-approaches`	A decision-ready comparison of viable technical approaches
`create-app-verifier`	A new project-local skill that safely drives and proves the real application
`maintain-app-verifier`	A source-and-live drift correction for an existing verifier and feature map
`test-engineer`	Focused test strategy, reproduction, implementation, and regression evidence
`performance-investigator`	Measurement-led performance diagnosis and before/after proof
`docs-maintainer`	Documentation aligned to authoritative current behavior, commands, and contracts
`refactor-maintainer`	Behavior-preserving structural change with an explicit invariant and complete caller migration

The selected skill reads the target repository’s instructions and uses its existing commands and conventions. Analysis-only skills stop before changing code; implementation skills preserve unrelated work and run focused plus repository-required validation.

Interfaces

The canonical interface is each skills/<name>/SKILL.md directory. Codex, Cursor, Agent Plugins-compatible hosts, and registry listings discover the same sources through thin metadata adapters. No HRA application, CLI, account, or service is required.

Evidence

Every skill states both a positive route and a negative boundary.
../../catalog/catalog.json records one owner and adjacent skills for every workflow.
../../catalog/routing-fixtures.json proves positive, negative, and pairwise routing cases.
Repository validation checks portable paths, host metadata, provenance, and the exact discovered skill inventory.

Third-party source history and licensing are recorded in PROVENANCE.md.

Boundaries

These skills do not coordinate an explicitly multi-phase delivery plan; use code-orchestrator for that result.
They do not implicitly apply a semantic transform to a business or writing question; invoke a named semantic-algos operator instead.
They do not replace repository instructions, application-specific policy, host scheduling, CI, review, merge, release, or deployment gates.
A real product regression remains product work; it is not verifier or documentation drift merely because a proof surface exposed it.

Questions

Which skill should I use to understand code?

Use repository-feature-map for broad navigation, semantic-code-analysis for one current code path, and reconstruct-rationale only when the question is historical intent.

Will an analysis skill implement its recommendation?

No. Analysis skills report evidence and stop. Use the appropriate implementation workflow only when the request includes the change.

Does the pack define my repository’s final gate?

No. The target repository’s instructions and delivery workflow remain authoritative.

Start

Review the skill that owns the requested result, then install hraness-engineering through one of the supported paths in the Hraness Agent Skills README.