vllnt/vstack
Evidence-led engineering workflows to plan, build, review, and ship software with AI agents. Part of the vllnt universe: open, sovereign tools for freedom by design (vllnt.com).
Draft or review framework-independent landing-page structure and conversion copy for one primary action.
Build and iterate a focused local prototype using observed behavior and human product decisions.
Deliver a scoped issue through a verified pull request and authorized merge to its target branch.
Review a pull request, coordinate authorized repairs and verification, then merge it when permitted.
Explain a real subject with a simple verified mental model and clear limits.
Sharpen an artifact or decision through focused evidence-based questions and concrete alternatives.
Simplify code and comments while preserving observable contracts and affected consumers.
Directly refactor source code and data paths toward measured reduction and efficiency while preserving behavior.
Speed up verified work for developers and agents by improving native build, lint, script, and CI workflows without weakening checks.
Audit the current machine on any OS for privacy leaks in identifiers, network traffic, telemetry, permissions, sharing, browsers, and file metadata, list every recommendation with its benefit, risk, and undo, then apply and verify only user-selected changes.
Audit the current machine on any OS for weak security settings, hidden or dormant processes, network exposure, and account risks, offer an approved local network check, then apply and verify only user-selected fixes.
Adapt an existing repository to applicable stack preferences through small verified stages.
Refocus a session at any point on its requested outcome, grounded decisions, simplicity, required quality, and efficient execution.
Improve a whole test suite by consolidating cross-file duplication and reducing execution cost while preserving coverage and defect detection.
Improve an interface’s hierarchy, consistency, responsiveness, and accessibility while preserving journeys.
Start a new project from answers, then install and verify it (@vllnt stack or a minimal baseline on another stack).
Diagnose a reported failure, isolate its cause, and apply authorized repairs with regression evidence.
Triage GitHub issues into an evidence-backed, atomic, actionable backlog and apply authorized changes.
Create, maintain, synchronize, or report a traceable roadmap that preserves direction and history.
Manage user, project, and nested instructions while preserving scope, precedence, and effective behavior.
Create, review, update, organize, or retire portable agent skills around a clear outcome and evidence.
Create, update, locate, or audit an evidence-backed project vision anchor and decision test.
Assess a codebase and prepare a source-backed, deduplicated issue backlog for human review.
Maintain a read-only planning session that returns an actionable plan without local or remote side effects.
Test a pull request in a full local or preview environment and return candidate-bound backend, frontend, and end-to-end QA evidence.
Assess an interface for evidence-backed usability, responsiveness, consistency, and accessibility findings without changing it.
Load at each session start to keep human collaboration proactive, concise, autonomous, and decision-ready.
Load at each session start to coordinate goal-driven work, useful delegation, convergence, and decisions.
Load at each session start to apply Vstack’s evidence-led, minimal reasoning principles throughout the work.