A dsh skill that gives Harness a local, file-based memory discipline (markdown plus index), installed via the dsh init helper; needs dsh 0.2.1+ and Python 3.9+.
DSH integration
Native runtime
Author-claimed
Safety audit
Unaudited
Last verified
2026-08-22
License
MIT
01What can it help you accomplish?
Give DeepSeek Harness a disciplined, local, file-based memory protocol (one fact per markdown file plus an always-loaded index)
A marked block in $DSH_HOME/AGENTS.md and a skill under <DSH_HOME>/skills/engramory, plus a separate .engramory-memory/ store the model follows as standing rules
DSH users who want a zero-infrastructure, human-readable agent memory discipline instead of a vector or graph DB
02How to install into DeepSeek Harness
Prerequisites
- Python 3.9+
- DeepSeek Harness dsh 0.2.1 or later
- The engramory repo cloned locally
Installation steps
- 01
Clone the repo and cd into it
- 02
Run `python tools/engramory_init.py dsh --install-skill` (defaults to $DSH_HOME; use --project-root for a project-local install)
- 03
Restart dsh; the protocol loads as standing rules via $DSH_HOME/AGENTS.md
Verify the integration
Not specified by the author
03DSH integration and capability boundaries
DSH integration via a skill installer: `python tools/engramory_init.py dsh --install-skill` writes an Engramory block into $DSH_HOME/AGENTS.md and installs the protocol under the dsh user skill root.
DSH skill install and AGENTS.md block
`python tools/engramory_init.py dsh --install-skill`→marked block in $DSH_HOME/AGENTS.md plus skill under <DSH_HOME>/skills/engramory
Local file-based memory store
agent notes or facts→markdown files in .engramory-memory/ plus an always-loaded index
04Who is it for? When not to use it?
Good for
- DSH users who want a zero-infrastructure, human-readable agent memory discipline instead of a vector or graph DB
Not for
- Install dsh 0.2.1 or later. On dsh 0.2.0 the skill installs but never activates (issue 8, older Cordis inject syntax); the preview-era cannot-install-third-party-plugins bug is fixed upstream in rc.7.
- The index cap is best-effort, not a global write guard: shell tools (Bash or PowerShell), MCP file tools, external editors, and sync clients bypass it. It is experimental (0.10.0); do not rely on it as a mandatory, cross-agent memory layer yet.
05Compatibility, maintenance and safety notes
- Install dsh 0.2.1 or later. On dsh 0.2.0 the skill installs but never activates (issue 8, older Cordis inject syntax); the preview-era cannot-install-third-party-plugins bug is fixed upstream in rc.7.
- The index cap is best-effort, not a global write guard: shell tools (Bash or PowerShell), MCP file tools, external editors, and sync clients bypass it. It is experimental (0.10.0); do not rely on it as a mandatory, cross-agent memory layer yet.
- Engramory is not a database, framework, or relevance-loaded skill - memory is plain-text markdown files you can open, edit, and diff; the live store is git-ignored.
MIT - actively maintained (latest release v0.10.0, 2026-08-20)
06Frequently asked questions
How do I install it for dsh?
Clone the repo, then run `python tools/engramory_init.py dsh --install-skill` (defaults to $DSH_HOME; add --project-root for a project-local install). Restart dsh so the block in $DSH_HOME/AGENTS.md loads.
Which dsh version is required?
dsh 0.2.1 or later. On 0.2.0 the skill installs but never activates (issue 8); the preview-era cannot-install-third-party-plugins bug is fixed in rc.7.
Is it a database?
No - memory is plain-text markdown files (one fact per file) plus one always-loaded index. No database, embeddings, or server; the live store is git-ignored.
Is the index cap a hard guard?
No - it is best-effort. Shell tools (Bash or PowerShell), MCP file tools, external editors, and sync clients bypass it. Engramory is experimental (0.10.0); do not treat it as a mandatory cross-agent memory layer yet.
07Related DSH workflows
memos
by memtensor
Self-evolving memory OS for LLM & AI Agents: ultra-persistent memory, hybrid-retrieval, and cross-task skill reuse, with 35.24% token savings and DeepSeek Harness support.
dsh-anchored-standard
by xiaobright
Two-phase DeepSeek Harness preset: Minimal-aligned bootstrap, then full Standard tools (Project2 98/99)
mem9
by mem9-ai
Unlimited memory for OpenClaw
mindmemos
by mindscale-noah
Gives dsh agents persistent cross-session memory: auto-recalls relevant user/task context before each turn and writes back lessons after each turn, evolving memory via schema learning and skill distillation
08Data and sources
An opinionated, zero-infrastructure memory *protocol* for small-scale, local, file-based agent memory
python tools/engramory_init.py dsh --install-skill
It writes a marked Engramory block into `$DSH_HOME/AGENTS.md` (dsh agent-instructions plugin)
Install **0.2.1 or later**: the preview-era cannot-install-third-party-plugins bug is fixed upstream (rc.7)
a separate `.engramory-memory/` store.
Memory is a folder of small, human-readable markdown files plus one always-loaded index.
This page is generated from the project’s public documentation, repository metadata and a structured parse of DSH Plugins; last verified on 2026-08-22. Found an error? Submit a correction.
Best DeepSeek Harness Plugins
Twelve plugins worth installing first — picked from the whole catalog, across every category.