Build minimal, sourced, freshness-aware context packets for project work.
npx skills add https://github.com/vTRKA/supervibe --skill context-engineering-vtrka
Please help me install this Agent Skill. Skill: context-engineering Source: https://github.com/vTRKA/supervibe/tree/main/skills/context-engineering Command: npx skills add https://github.com/vTRKA/supervibe --skill context-engineering-vtrka
It prevents AI coding sessions from becoming unfocused by constructing a compact, freshness-aware context packet with provenance, so decisions and edits rely on current evidence rather than memory or guesswork.
Ask your AI to prepare context for a non-trivial change by requesting a minimal, sourced, freshness-aware context packet suitable for making the next safe edit.
High-intent search queries and answers about installing and using this skill.
Build sourced context by constructing a compact, freshness-aware context packet with provenance to ensure decisions rely on current evidence rather than memory. It selects, structures, and compresses only the evidence needed for planning, implementation, and review.
Context engineering is the process of building minimal, sourced task context for AI work. It enforces evidence prioritization from user requests to local sources, separates facts from assumptions with confidence boundaries, and emits structured packets for verification.
Separate facts from assumptions during agent handoff by using confidence boundaries and freshness diagnostics within the context packet. The system records what is verified fact versus assumption and caps confidence when evidence is partial or indices are stale.
Yes, this approach works with project memory and code search by applying scope-safe routing to the right sources. It uses CodeGraph, runtime receipts, and local sources to avoid irrelevant context dumps and ensure the next agent can act confidently.
You need freshness-aware context packets for code reviews when making reliable decisions requires current evidence. The packets include freshness diagnostics and track uncertainty, ensuring edits do not rely on stale indices or guesswork during multi-file workflows.
lyq-linQuery a local knowledge base for Ant Design component semantics and documentation links.
oikon48Launch four Explore agents in parallel to investigate distinct codebase areas.
asaeedAutomate routing, task scanning, and memory updates for Obsidian vault notes.
jmanhypeRetrieve relevant bullets and apply incremental deltas to refine agent guidance.
