dsh-relay

Overview Schema Source Related Servers Discussions Admin

Steer a Harness run

steer_run

Destructive

Insert a correction into an active run by specifying the run ID, task parameters, and content. This tool fixes issues in ongoing runs without replacing the entire run, using path references to guide the harness.

Instructions

Durably insert a correction into an active run. Task parameters are path-reference-only: provide the authorized workspace, reviewTargets, contextReadScope, excludedPaths, writeScope, and acceptance criteria. reviewTargets identify what to assess; they are not a read whitelist. contextReadScope declares where Harness may search and read supporting implementation, tests, configuration, and architecture material. Never embed source text, diffs, file dumps, encoded source, or repository archives. Harness reads named files from the authorized workspace itself. These fields are task instructions, not enforced per-path filesystem isolation. The selected Harness model may send content it reads to its configured model provider even though Relay itself uses a loopback address.

Input Schema

TableJSON Schema

Name	Required	Description
`task`	No	Task parameters are path-reference-only: provide the authorized workspace, reviewTargets, contextReadScope, excludedPaths, writeScope, and acceptance criteria. reviewTargets identify what to assess; they are not a read whitelist. contextReadScope declares where Harness may search and read supporting implementation, tests, configuration, and architecture material. Never embed source text, diffs, file dumps, encoded source, or repository archives. Harness reads named files from the authorized workspace itself. These fields are task instructions, not enforced per-path filesystem isolation. The selected Harness model may send content it reads to its configured model provider even though Relay itself uses a loopback address.
`runId`	Yes
`content`	No	Task parameters are path-reference-only: provide the authorized workspace, reviewTargets, contextReadScope, excludedPaths, writeScope, and acceptance criteria. reviewTargets identify what to assess; they are not a read whitelist. contextReadScope declares where Harness may search and read supporting implementation, tests, configuration, and architecture material. Never embed source text, diffs, file dumps, encoded source, or repository archives. Harness reads named files from the authorized workspace itself. These fields are task instructions, not enforced per-path filesystem isolation. The selected Harness model may send content it reads to its configured model provider even though Relay itself uses a loopback address.
`idempotencyKey`	No

Schema Changelog

Changes observed during successful MCP inspections.

Changed3 schema fields changedv0.2.13Sep 10, 2026

changedInput schema / properties / content / description

Previous value: -"Task parameters are path-reference-only: provide the authorized workspace, file or directory locations, scope, and acceptance criteria. Never embed source text, diffs, file dumps, encoded source, or repository archives. Harness reads named files from the authorized workspace itself. This rule is identical for read-only and write-capable permissions."New value: +"Task parameters are path-reference-only: provide the authorized workspace, reviewTargets, contextReadScope, excludedPaths, writeScope, and acceptance criteria. reviewTargets identify what to assess; they are not a read whitelist. contextReadScope declares where Harness may search and read supporting implementation, tests, configuration, and architecture material. Never embed source text, diffs, file dumps, encoded source, or repository archives. Harness reads named files from the authorized workspace itself. These fields are task instructions, not enforced per-path filesystem isolation. The selected Harness model may send content it reads to its configured model provider even though Relay itself uses a loopback address."

changedInput schema / properties / content / items / oneOf

Previous value: -[
-  {
-    "properties": {
-      "text": {
-        "description": "Task parameters are path-reference-only: provide the authorized workspace, file or directory locations, scope, and acceptance criteria. Never embed source text, diffs, file dumps, encoded source, or repository archives. Harness reads named files from the authorized workspace itself. This rule is identical for read-only and write-capable permissions.",
-        "maxLength": 100000,
-        "type": "string"
-      },
-      "type": {
-        "const": "text",
-        "type": "string"
-      }
-    },
-    "required": [
-      "type",
-      "text"
-    ],
-    "type": "object"
-  },
-  {
-    "properties": {
-      "data": {
-        "minLength": 1,
-        "type": "string"
-      },
-      "mediaType": {
-        "enum": [
-          "image/png",
-          "image/jpeg",
-          "image/webp",
-          "image/gif"
-        ],
-        "type": "string"
-      },
-      "name": {
-        "type": "string"
-      },
-      "type": {
-        "const": "image",
-        "type": "string"
-      }
-    },
-    "required": [
-      "type",
-      "mediaType",
-      "data"
-    ],
-    "type": "object"
-  }
-]New value: +[
+  {
+    "properties": {
+      "text": {
+        "description": "Task parameters are path-reference-only: provide the authorized workspace, reviewTargets, contextReadScope, excludedPaths, writeScope, and acceptance criteria. reviewTargets identify what to assess; they are not a read whitelist. contextReadScope declares where Harness may search and read supporting implementation, tests, configuration, and architecture material. Never embed source text, diffs, file dumps, encoded source, or repository archives. Harness reads named files from the authorized workspace itself. These fields are task instructions, not enforced per-path filesystem isolation. The selected Harness model may send content it reads to its configured model provider even though Relay itself uses a loopback address.",
+        "maxLength": 100000,
+        "type": "string"
+      },
+      "type": {
+        "const": "text",
+        "type": "string"
+      }
+    },
+    "required": [
+      "type",
+      "text"
+    ],
+    "type": "object"
+  },
+  {
+    "properties": {
+      "data": {
+        "minLength": 1,
+        "type": "string"
+      },
+      "mediaType": {
+        "enum": [
+          "image/png",
+          "image/jpeg",
+          "image/webp",
+          "image/gif"
+        ],
+        "type": "string"
+      },
+      "name": {
+        "type": "string"
+      },
+      "type": {
+        "const": "image",
+        "type": "string"
+      }
+    },
+    "required": [
+      "type",
+      "mediaType",
+      "data"
+    ],
+    "type": "object"
+  }
+]

changedInput schema / properties / task / description

Previous value: -"Task parameters are path-reference-only: provide the authorized workspace, file or directory locations, scope, and acceptance criteria. Never embed source text, diffs, file dumps, encoded source, or repository archives. Harness reads named files from the authorized workspace itself. This rule is identical for read-only and write-capable permissions."New value: +"Task parameters are path-reference-only: provide the authorized workspace, reviewTargets, contextReadScope, excludedPaths, writeScope, and acceptance criteria. reviewTargets identify what to assess; they are not a read whitelist. contextReadScope declares where Harness may search and read supporting implementation, tests, configuration, and architecture material. Never embed source text, diffs, file dumps, encoded source, or repository archives. Harness reads named files from the authorized workspace itself. These fields are task instructions, not enforced per-path filesystem isolation. The selected Harness model may send content it reads to its configured model provider even though Relay itself uses a loopback address."

Changed3 schema fields changedv0.2.9Sep 1, 2026

addedInput schema / properties / content / description

Added value: +"Task parameters are path-reference-only: provide the authorized workspace, file or directory locations, scope, and acceptance criteria. Never embed source text, diffs, file dumps, encoded source, or repository archives. Harness reads named files from the authorized workspace itself. This rule is identical for read-only and write-capable permissions."

changedInput schema / properties / content / items / oneOf

Previous value: -[
-  {
-    "properties": {
-      "text": {
-        "maxLength": 100000,
-        "type": "string"
-      },
-      "type": {
-        "const": "text",
-        "type": "string"
-      }
-    },
-    "required": [
-      "type",
-      "text"
-    ],
-    "type": "object"
-  },
-  {
-    "properties": {
-      "data": {
-        "minLength": 1,
-        "type": "string"
-      },
-      "mediaType": {
-        "enum": [
-          "image/png",
-          "image/jpeg",
-          "image/webp",
-          "image/gif"
-        ],
-        "type": "string"
-      },
-      "name": {
-        "type": "string"
-      },
-      "type": {
-        "const": "image",
-        "type": "string"
-      }
-    },
-    "required": [
-      "type",
-      "mediaType",
-      "data"
-    ],
-    "type": "object"
-  }
-]New value: +[
+  {
+    "properties": {
+      "text": {
+        "description": "Task parameters are path-reference-only: provide the authorized workspace, file or directory locations, scope, and acceptance criteria. Never embed source text, diffs, file dumps, encoded source, or repository archives. Harness reads named files from the authorized workspace itself. This rule is identical for read-only and write-capable permissions.",
+        "maxLength": 100000,
+        "type": "string"
+      },
+      "type": {
+        "const": "text",
+        "type": "string"
+      }
+    },
+    "required": [
+      "type",
+      "text"
+    ],
+    "type": "object"
+  },
+  {
+    "properties": {
+      "data": {
+        "minLength": 1,
+        "type": "string"
+      },
+      "mediaType": {
+        "enum": [
+          "image/png",
+          "image/jpeg",
+          "image/webp",
+          "image/gif"
+        ],
+        "type": "string"
+      },
+      "name": {
+        "type": "string"
+      },
+      "type": {
+        "const": "image",
+        "type": "string"
+      }
+    },
+    "required": [
+      "type",
+      "mediaType",
+      "data"
+    ],
+    "type": "object"
+  }
+]

addedInput schema / properties / task / description

Added value: +"Task parameters are path-reference-only: provide the authorized workspace, file or directory locations, scope, and acceptance criteria. Never embed source text, diffs, file dumps, encoded source, or repository archives. Harness reads named files from the authorized workspace itself. This rule is identical for read-only and write-capable permissions."

First observedv0.2.6Aug 30, 2026

TDQS

A4.3/5.0

Behavior5/5

Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?

The description adds substantial behavioral detail beyond the annotations: task parameters are path-reference-only, no source text may be embedded, Harness reads named files itself, fields are not enforced filesystem isolation, and the model may send read content to its configured provider despite Relay's loopback address. These are meaningful privacy and execution-model disclosures that a caller needs.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Conciseness4/5

Is the description appropriately sized, front-loaded, and free of redundancy?

The description is front-loaded with the essential purpose and then provides dense, valuable warnings. Each sentence adds a distinct constraint or behavioral fact; the only redundancy is that the same text appears verbatim in the schema's parameter descriptions, but the tool description itself is not padded.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Completeness4/5

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

For a destructive, non-idempotent mutation on an active run, the description covers the critical safety and privacy aspects, including what must not be included in the payload and how Harness accesses files. It does not explain the return value or success/error behavior, but with no output schema and a clear purpose this is a minor gap.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Parameters3/5

Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?

Schema coverage is only 50%: runId and idempotencyKey have no schema descriptions, and the tool description does not compensate for those. However, the description does clarify the task/content parameter semantics thoroughly, explaining path-reference usage, reviewTargets, contextReadScope, and the prohibition on embedding source material, so it partially makes up for the gaps.

Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.

Purpose5/5

Does the description clearly state what the tool does and how it differs from similar tools?

The opening clause, 'Durably insert a correction into an active run,' uses a specific verb, resource, and state ('active run'), which clearly distinguishes it from siblings like start_run, reply_run, and cancel_run. It also names the core operation type ('correction') rather than merely restating the tool's name.

Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.

Usage Guidelines4/5

Does the description explain when to use this tool, when not to, or what alternatives exist?

The description gives clear context: it is for inserting a correction into an already active run, which implies it should not be used for starting, stopping, or merely observing runs. It does not explicitly name alternatives or exclusion conditions, so it misses the top score, but the context is unambiguous.

Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.

Steer a Harness run

Instructions

Input Schema

Schema Changelog

TDQS

Other Tools