Engram memory
Give the agent durable, local memory with engram — recall past decisions before answering, and persist new decisions, preferences and facts as they happen. Use when work spans sessions or the user says "remember".From its SKILL.md
npx -y skills add jnMetaCode/skillet --skill engram-memoryAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
2 things to look at
- 1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
- runs commandsInstructs the agent to run 6 commands, including `npx @jnmetacode/engram serve &` and 5 more.
What its file declares
Copied from the file, not written here
The file declares its own license as MIT. That is the author’s claim about this one file, and it is not the same thing as the license GitHub reports for the repository, which is listed with the other numbers below.
SKILL.md
2.8 KB, 625 tokens by cl100k_base, as published. Nobody here has run it
engram-memory
You have access to a local, private memory layer (engram). Treat it as your long-term memory: read it before you answer, write to it when something durable happens. Everything stays on the user's machine.
Setup (once)
If the engram_recall / engram_remember MCP tools are available, use those.
Otherwise use the CLI (zero install):
npx @jnmetacode/engram serve & # HTTP API on :7077, or
npx @jnmetacode/engram mcp # as an MCP server
When to RECALL
Before answering anything that may depend on prior context, search memory first:
- "What did we decide about X?" / "Why did we choose Y?"
- Resuming work after a gap ("where were we?")
- Anything referencing a person, project, deadline, or preference you don't see in the current conversation.
npx @jnmetacode/engram recall "pricing decision" --since month
Quote the recalled passage with its citation (file/date) rather than paraphrasing from your own context — the citation is the point.
When to REMEMBER
Persist a memory when the conversation produces something with a shelf life:
- A decision and its why ("we picked Postgres over SQLite because …")
- A user preference ("always use pnpm", "no AI attribution in commits")
- A fact that took effort to establish (a root cause, a benchmark number)
npx @jnmetacode/engram remember "2026-06-10: chose scoped npm names (@org/pkg) because unscoped were taken"
Rules for good memories:
- One fact per memory. Atomic entries rank and recall better.
- Date it. Lead with an absolute date — engram's ranking is time-aware.
- Include the why, not just the what; the why is what future-you needs.
- Don't store what the repo already records (code, git history, docs).
REINFORCE what proved right (self-improving recall)
When a recalled memory turned out to be the correct answer — the fix worked, the user confirmed — say so, and similar future queries will rank that source higher:
npx @jnmetacode/engram reinforce "staging deploy fails" deploy-notes
(MCP: the engram_reinforce tool, {query, source}.) Only reinforce
verified answers; reinforcing guesses trains the memory to be confidently
wrong. For the full improvement loop, see the self-evolve skill.
Verify it stuck
After remembering, do a quick recall of a keyword from the new memory. If it
doesn't come back first, rewrite it more concretely.
What ships with it
Read from the repository
Just SKILL.md. No reference files, no scripts.
Gives 0 of the 12 instructions most memory context skills give in 625 tokens
Counted across 754 of the 1,056 authors here whose files we hold, read 2026-09-06
- Preserve existing content structurein 15 of 754, across 9 files
- Front-load the leading wordin 14 of 754, across 10 files
- Update existing entries instead of duplicatingin 14 of 754, across 7 files
- Keep CLAUDE.md under one hundred linesin 14 of 754, across 12 files
- Read CLAUDE.md at the project rootin 14 of 754
- Keep each meaning in a single source of truthin 12 of 754, across 8 files
- Redact sensitive information before committingin 11 of 754, across 4 files
- Scan for all CLAUDE.md filesin 11 of 754, across 7 files
- Use frontmatter for metadata on filesin 10 of 754, across 3 files
- Repeat user interactions 10 timesin 10 of 754, across 4 files
- Write the CLAUDE.md file into the target folderin 10 of 754, across 8 files
- Use memlab to process snapshotsin 9 of 754, across 3 files
Said here and by no other author read
- Persist durable decisions and preferences
- Quote recalled passages with citations
- Lead memories with an absolute date
- Reinforce verified recalled memories
Grouped from the skills themselves: near-identical wordings counted once, and counted by distinct author, so one author publishing three of these counts once. Length counted with cl100k_base; the agent that loads this file may tokenize it differently.