DSH Session Insights — DeepSeek Harness Retrospectives

    1

    Turn DeepSeek Harness sessions into a local HTML workflow review: projects, tool failures, retries and improvements. Requires DSH.

    Free

    1 useSecurity scanned

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    DSH Session Insights — DeepSeek Harness Retrospectives

    Example session with this skill installed

    Review my last 30 days of DeepSeek Harness work and help me understand project patterns, tool failures and retries.

    • Read your context and instructions
    • Compiled the dsh session insights

    The download includes an English dashboard screenshot generated from fully synthetic test sessions. It illustrates the report layout; it is not a transcript or an analysis of your sessions. Your run writes its own local HTML dashboard and companion JSON.

    Connects securely to your tools. The creator never sees your data.

    About this skill

    Review your DeepSeek Harness workflow

    Run /session-insights to turn DSH session history into a local HTML dashboard and companion JSON. Review project and workflow patterns, tool failures, retries, completion evidence, and recommendations tied to the selected sessions.

    Requirements and installation

    Requires DeepSeek Harness >=0.2.0-rc.2 and Node.js ^22.19.0 or >=24.0.0. The native Bundle needs DSH's commands, tools and sessionQuery services. It does not require Python.

    For DSH Web, install with dsh plugin --profile web add dsh-session-insights@0.5.2, start DSH Web, then enter /session-insights --days 30 --locale en in the DSH composer. Official Desktop uses its own plugin controls or bundled command carrier; see the source README.

    This ZIP preserves the published Bundle source and adds a marketplace entry SKILL.md. Follow the DSH installation route; dropping SKILL.md into another agent does not install DSH or provide its session services. Other agents' logs are outside this tool's scope.

    Privacy and model usage

    Add --deterministic for offline analysis. The default model-assisted workflow sends bounded, sanitized evidence through the current DSH agent to its configured model provider; that provider's data handling and usage charges apply. Redaction reduces disclosure but is not a guarantee that evidence contains no secrets. Reports and companion JSON remain local.

    Free and open source

    Version 0.5.2, MIT licensed. The ZIP includes LICENSE. Public source, complete installation instructions and future releases: https://github.com/GreenLv/dsh-session-insights

    The dashboard screenshot included in the ZIP comes from fully synthetic test data, not a private user's sessions. This is a workflow retrospective, not live telemetry, token billing or a quality score for your work.

    Requires DSH session services; does not analyze non-DSH logs. Generic marketplace installation hints do not replace DSH Bundle installation. Token totals are not billing. Semantic analysis uses the configured model provider; deterministic mode is offline. Analysis is bounded to 2,000 snapshots and 64 MiB of serialized input per run. The optional Python CLI and its managed skill require a separate installation.

    How to use

    1. 1

      Connect your agent to the Agensi MCP

      Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup

    2. 2

      Sign in to Agensi

      Your agent shows a sign-in link. Sign in and click Allow.

    3. 3

      Ask for what you need

      Your agent loads this skill and uses it on your task. It always asks before any payment.

    Reviews

    No reviews yet

    Be one of the first to try it. Every listed skill goes through the trust checks below.

    Security scanned

    Checked by our 8-point scan before listing

    1 use

    Times used to date

    Free forever

    No account required to browse

    Trust & safety

    Security scan

    Security scan passed 1 month ago

    • Free with an Agensi account

    Requires

    Npmjs
    Node.js
    Python

    Permissions

    Terminal / Shell
    Read Files
    Write Files
    Network Access
    Environment Variables

    Allowed Hosts

    registry.npmjs.org

    Installation uses the DSH plugin manager and npm registry. Native analysis reads selected DSH sessionQuery snapshots and writes marked reports under DSH_HOME/insights/runs; DSH_HOME is used only to locate local storage. Deterministic analysis is offline. Semantic analysis uses the current DSH agent's configured provider; this package collects no separate API key. The optional Python CLI is a distinct route.

    Listed1 month ago

    What's inside

    Frequently Asked Questions

    It reviews selected DeepSeek Harness session history and writes a local HTML workflow dashboard and companion JSON. The report covers project and workflow patterns, tool failures, retries, completion evidence and recommendations grounded in the selected sessions.

    Requires DeepSeek Harness >=0.2.0-rc.2 and Node.js ^22.19.0 or >=24.0.0. Install the DSH Bundle using the host-specific route in the GitHub README. A generic agent's skills folder does not provide DSH's commands, tools or sessionQuery services, and this tool does not analyze other agents' logs.

    Use --deterministic for offline analysis. The default model-assisted workflow sends bounded, sanitized evidence through the current DSH agent to its configured model provider; that provider's data handling and usage charges apply. Redaction reduces disclosure but does not guarantee that evidence contains no secrets. Reports and companion JSON remain local.

    No. Analysis is scoped to the selected DSH sessions and is bounded to 2,000 snapshots and 64 MiB of serialized input per run. It is a workflow retrospective, not live telemetry, token billing or a quality score. Read the report's coverage and limitations when interpreting its findings.

    Yes. The MIT-licensed ZIP contains the published dsh-session-insights 0.5.2 Bundle source, LICENSE, README documentation, a marketplace SKILL.md entry and an English dashboard screenshot generated from fully synthetic test data. Source and future releases are available at https://github.com/GreenLv/dsh-session-insights.

    New in LLM Ops