Skill profile
continual-learning
Automatically and incrementally keeps AGENTS.md up to date, mining new session transcripts on a cadence and updating matching bullets in place rather than rewriting the file.
What it actually does
- Combines a `stop` hook, a `continual-learning` skill, and an `agents-memory-updater` subagent to mine new or changed session transcripts and update `AGENTS.md` incrementally — reading the existing file first and updating matching bullets in place rather than rewriting it.
When to use it — and when not to
Use it when
- You want AGENTS.md to stay current with what the agent actually learns across sessions, without manually curating it.
Not for
- You want it to fire mid-conversation on demand — `disable-model-invocation: true` means it never self-triggers; only the `stop` hook cadence (10 turns / 120 minutes by default) invokes it.
Agent compatibility
- Cursor
Cursor-native — built on Cursor's stop hooks, subagents, and local hook-state files.
Confirmed means the skill’s own source or vendor states it works with that agent. Portable means it only reaches that agent through the shared SKILL.md format and a directory convention — real, but a weaker claim. An agent with no entry here isn’t known to be unsupported; it simply isn’t documented either way, so nothing is claimed. Full model on Agent Skills Compatibility.
Requirements and operational surface
Depends on
- Cursor `stop` hooks
- Cursor subagents (`agents-memory-updater`)
What it actually does at runtime
- Reads session transcripts
- Writes incremental edits to `AGENTS.md`
- Keeps local hook state under `.cursor/hooks/state/`
Evidence and provenance
Maintained by Cursor (official). Primary source: cursor/plugins — continual-learning, last checked against this page’s claims on . This page summarizes and analyzes that source — it isn’t a copy of its SKILL.md or README, and installation happens at the source, not here.