AI Agent Hub
Back to plugins
dsh-experience-map preview

dsh-experience-map

Memory Updated 2026.09.14

Run the following command in DeepSeek Harness:

dsh plugin install alcheme-labs/dsh-experience-map

Paste the following prompt into your AI chat to install this plugin:

Run dsh plugin install alcheme-labs/dsh-experience-map in the DeepSeek Harness terminal to install this plugin; the full source code is available at https://github.com/alcheme-labs/dsh-experience-map

About this plugin

Ordinary chat history reminds a model what was said, but it cannot reliably tell it which steps actually worked, under which conditions, who approved them, or whether the old result is still valid. Vector search retrieves similar text, yet similarity alone is not permission to reuse a procedure. dsh-experience-map turns real session evidence from completed tasks into typed, versioned, evidence-graded Experiences, and re-validates applicability in the current environment before every reuse, preventing mistaken workflows from being silently repeated.

At its core, the plugin introduces a full lifecycle: automatic suggestion detection, a user-controlled save gate, deterministic hard-gate matching with optional local semantic ranking, a Preflight applicability check, an explicit Plan approval step, current verification, an immutable Settlement, and the ability to publish a new Version or Forget future retrieval. Six Experience kinds—Procedure, Diagnostic, Strategy, Preference Policy, Fact, and Causal—each carry their own evidence rules, so a causal hypothesis is never silently promoted to established causality. SQLite owns all durable records; Markdown views and relation maps are readable projections, not competing stores.

Screenshots

Use Cases

  • Reuse an approved workflow when a similar task recurs instead of re-exploring every time
  • Accumulate verifiable success paths across sessions so mistaken steps are not silently repeated
  • Run experience recall and governance headlessly in CI pipelines without a browser
  • Gate high-stakes actions behind a Preflight check and explicit Plan approval to prevent blind reuse of stale solutions

Best For

  • Developers who drive DeepSeek Harness through complex, multi-step engineering tasks
  • Platform or ops teams that must approve and audit AI-generated knowledge
  • Engineers who run Harness inside agent swarms or automated pipelines