The Engramory curated-memory discipline as an installable plugin ([npm: dsh-engramory](https://www.npmjs.com/package/dsh-engramory)): a deterministic 200-line / 25 KB cap on the `MEMORY.md` index via `ctx.tools.guard()` — growth denied, a shrinking rewrite always passes — plus the protocol registered as a runtime skill. The store is plain markdown, one file per fact, shared with Claude Code, Codex, Kiro, and OpenClaw.
Install
# from npm (prebuilt)
dsh plugin --profile web add dsh-engramory
# from GitHub (first run asks for allowBuilds approval — follow the hint, retry)
dsh plugin --profile web add github:tinqiao-oss/engramory#path:/adapters/dsh/plugin
Any plugin you install runs third-party code with your own permissions — it can read your files, use your credentials, and reach the network, and tool approvals don’t sandbox it. GitHub-sourced plugins also run build scripts at install time. Only install sources you trust, and pin a commit (github:owner/repo#sha).
README
Curated, file-based long-term memory for DeepSeek Harness — plain markdown notes you can open with anything, one store shared across every host you use, and an index cap that is a real refusal rather than a request. (If the store lives inside a project's git repo it must be git-ignored — memories carry machine-local detail; see SKILL.md §1. A dedicated private repo for the store itself is fine.)
Part of Engramory.
Why a plugin, when an AGENTS.md block already works
The block does carry the discipline, and for recall that is enough. Two things it cannot do:
The cap becomes deterministic. Engramory keeps its index under 200 lines / 25 KB
because the index loads every session and everything past the cap silently stops being
recalled. On most hosts that limit is rules plus a checker the agent has to remember to
run. dsh exposes ctx.tools.guard() — a synchronous, monotonic refusal: once a guard
returns a reason, no later listener can turn it back into an allow. So here the limit is
enforced, not requested — once the plugin is actually installed and running; a
profile install currently fails upstream (see "Known limits"), and until then the cap on
dsh is still rules plus the checker. Outside Claude Code, dsh is the first host with the
shim actually written (a few others expose equivalent pre-write seams — see PORTING.md —
but have no shim yet).
The protocol arrives with the plugin. ctx.skills.register() contributes the skill at
runtime, so it does not depend on landing files in one of the five skill roots dsh scans —
a path that is easy to get subtly wrong and fails silently when you do.
Install
dsh plugin --profile <name> add dsh-engramory
That is the whole install. The package ships a dsh.bundle manifest pointing at its own
cordis.patch.yml, so the row is inserted into the profile's plugin tree for you — no
hand-editing. To change a default, patch the EXISTING row by id in your profile's own
patch layer — do not use a second - insert: (insert always appends, so you would
end up with two engramory rows and the original caps still enforced):
- id: engramory
config:
indexName: MEMORY.md
maxLines: 200
maxBytes: 25600
A patch replaces the targeted row's whole config, so list every key you mean to keep.
For the store itself and the always-on block, use the installer in the Engramory repo
(python tools/engramory_init.py dsh --install-skill). This plugin enforces the cap and
supplies the protocol; it does not create the store.
Configuration
| Field | Default | Meaning |
|---|---|---|
indexName |
MEMORY.md |
Basename treated as the memory index (case-insensitive; both file_path and str_replace_editor's path are checked). An empty or non-string value falls back to the default. Nothing else is inspected. |
maxLines |
200 |
Hard line cap. Non-positive or non-finite values fall back to the default. |
maxBytes |
25600 |
Hard UTF-8 byte cap (25 KB). |
registerSkill |
true |
Set false to keep the cap but skip the runtime skill. |
skill |
built-in | Replace the skill body with your own markdown. |
What the guard actually does
| Call | Decision |
|---|---|
write (or str_replace_editor create) ending within caps |
allow |
write/create that GROWS the index past a cap |
deny, naming the numbers and what to compact |
write/create that SHRINKS or keeps an over-cap index |
allow — incremental compaction (210 → 205 → 198) must stay possible |
Edit carrying old_str/new_str: result simulated |
judged by the RESULT, same grow/shrink rule as a write |
Unsimulable partial (e.g. insert) on an over-cap index |
deny, telling the agent a shrinking whole-file write passes |
| Unsimulable partial on a healthy index | allow |
read, view, or any unknown tool |
allow — recall must never be blocked, even over cap |
| Any write to any other file | allow — not the guard's business |
| Missing/unreadable current index on a whole-file write | treated as EMPTY (mirrors the Python guard) — a within-caps write passes, an over-cap first write is refused |
| Missing/unreadable index on a partial edit | allow — a guard must never block work over a path it cannot read |
| Malformed execution (no arguments, non-string path/content) | allow — not recognisably a write |
The deny rule mirrors hooks/engramory_index_guard.py exactly: refuse only a result
that is over a cap AND grew past the current file. The line count ignores a trailing
newline, so an index sitting exactly at the cap stays writable.
Known limits
- An unsimulable partial that crosses the cap from under it is not caught.
insertand friends do not carry the resulting text. The breach is caught by the next whole-file write, or byengramory_check.py. What is guaranteed: an already-over index cannot be grown further, and a shrinking write always passes. - The tool roster follows dsh's documented tool-fs contract (
write/editwithfile_path,str_replace_editorwithpath) and is deliberately conservative: unknown tools pass. Until the plugin runs inside a real profile this mapping is documentation-verified only. - Not yet verified end to end inside dsh. The guard's decision table is covered by
node --test(21 cases, run in Engramory's CI), and the wiring for the AGENTS.md block and skill discovery was dogfooded against a livedeepseek-v4-flashsession. But installing a third-party plugin into a profile currently fails on dsh 0.1.0-rc.6:dsh pluginshells out topnpm(not bundled), and a directpnpm addthen dies onERR_PNPM_FETCH_404for@deepseek-ai/dsh-type-meta, a package the published tree depends on but which is not on the registry. That is a preview-packaging problem upstream, not a plugin defect; this note goes away once a profile install resolves. - dsh is a developer preview and its plugin API can change. This plugin deliberately
touches only
ctx.tools.guard()andctx.skills.register()so it stays cheap to fix.
License
MIT — see the Engramory repository.
Links
More in this category
volcengine/OpenViking#examples/dsh-memory-plugin★ 28525
OpenViking memory and context bundle for DeepSeek Harness: pre-step auto-recall and profile injection, session capture, `viking://` URI guarding, and recall/write memory tools backed by an OpenViking server.
vectorize-io/hindsight#coding-agents★ 20019
Hindsight, agent memory that learns: long-term project memory with auto recall and retain, knowledge pages, deep reflection, and per-repo memory banks.
omdsh-dev/dsh-mnemon★ 40
Cross-agent, local-first persistent memory plugin for DeepSeek Harness (DSH), powered by Mnemon. It shares long-term memory across Mnemon-enabled agents and adds runtime memory, searchable project documents, semantic recall, knowledge graph, and a Sidebar UI.
LoserFox/distill★ 19
Automatic conversation distillation: background subagent reflection + skill create/update.
Tyan66666/billion-context-dsh★ 16
Model-driven context compression (Active Context Pruning) for DeepSeek Harness: the model decides when and what to compress.
Aik358/dsh-auto-memory★ 15
Auto-memory for DSH: three-layer memory (user / project notes / daily logs) with automatic injection, per-turn auto-consolidation, AI greetings, smart search, a calendar view and a settings page, plus inheritance of other AI tools' memories.