AI Agent Hub
Back to plugins
🧰

dsh-http-probe

Web Tools Updated 2026.08.15

Run the following command in DeepSeek Harness:

dsh plugin install TwistedRiCen/dsh-http-probe

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

Run dsh plugin install TwistedRiCen/dsh-http-probe in the DeepSeek Harness terminal to install this plugin; full source is available at https://github.com/TwistedRiCen/dsh-http-probe

About this plugin

In the DeepSeek Harness ecosystem, models often need to run lightweight probes against deployment-side services to confirm endpoint reachability or inspect status, but exposing arbitrary URL access would introduce a clear security risk. dsh-http-probe closes that gap by offering a tightly scoped probe: the model may only supply a URL path, while the host, port, and protocol are locked down by deployment configuration, effectively preventing any request to an unauthorised destination.

The execution boundaries are equally well-defined. Response body reads are capped at a configurable byte limit (default 8,192 bytes); once the cap is hit, reading stops and the structured result carries truncated: true. A cooperative timeout budget is declared on the tool and enforced by the host tool-call timeout policy. Redirects are never followed automatically. The canonical result is always a structured object with status code, Content-Type, body, and truncation flag, making downstream logic deterministic. The path parameter is strictly validated: it must begin with a single slash and may not contain path traversal segments, query strings, fragments, or control characters.

The plugin is ideal for developers building on DeepSeek Harness who want their model to read health-check or endpoint status from its own deployment, particularly in teams with a hard requirement that the model must not initiate arbitrary outbound requests. Note that the plugin requires an exact host dependency on a specific rc build of dsh-tools that is not yet published on the public npm registry, so installing from the prebuilt GitHub Release tarball is the recommended path.

Use Cases

  • Model verifies whether a deployment-owned health endpoint is reachable and reads its status code
  • Let a model access only a fixed API path in a strictly restricted environment
  • Build deterministic network probe and response validation logic in Harness workflows

Best For

  • Developers building applications on DeepSeek Harness
  • Teams with security compliance requirements on model outbound access
  • Engineers who need their model to sense deployment service state