Agent Skill for Codex, Cursor, and Claude
Give every AI session the context it is missing.
Syncora gives coding agents durable, local project memory without an extension, hosted service, or background process.
One skill. Three agent environments.
Durable project memory, right where you already work.
00 / The Problem
Important context doesn't vanish.It decays into residue.
Large context windows preserve traces, not working clarity. As conversations grow, decisions compress into shorter summaries, losing rationale, scope, and rejected alternatives until they no longer map cleanly to the code they govern.
"Syncora is an open-source Agent Skill for Codex, Cursor, and Claude. It loads bounded project context only when relevant, checks every initialized request for durable outcomes before the reply ends, and updates one canonical project or workstream hub instead of creating competing notes. It runs only during the active request, with no extension, server, or daemon."
"Syncora gives coding agents durable local memory. It loads only relevant context, saves useful decisions automatically before replying, and keeps current truth in project and workstream hubs. Changed files can flag possible drift, but cannot invent new rationale."
"Syncora stores durable decisions in local project hubs and brings back relevant context for new tasks. It checks changed sources for possible stale memory."
"Syncora stores important chat context for code and retrieves it when that code changes. It links decisions to code anchors and runs mostly in the background."
"Syncora stores context for code."
Important decisions do not disappear. They decay into unusable context.
The model may retain the topic.It loses the operating definition.
Syncora context pack