Skip to content

Latest commit

 

History

History
71 lines (57 loc) · 8.51 KB

harness-capabilities.md

File metadata and controls

71 lines (57 loc) · 8.51 KB
title Harness capabilities

This page lists what each Harness supports on each placement. Core decides admission from the Harness's declaration in internal/harnessconfig/<package>/declaration.go, and the Runtime that runs the Session must also advertise the operation. The linked contracts define each operation; Harness onboarding describes how a Harness is qualified.

Anything this page does not list follows the Harness's native behavior with the selected model, as in a local run. Each adapter guide lists the native defaults it overrides for unattended cloud execution (design rule).

Status Meaning
Verified Core admits it, and real-model acceptance through the pinned official client passed on that placement
Admitted Core admits it through the same Runtime path, but no real-model acceptance has run on that placement
Rejected Core rejects the request before execution

Placements are none (no Environment), hosted (openai_hosted) and self-hosted (self_hosted). microsandbox, E2B and OpenSandbox run the same Runtime and adapters, and every Verified hosted cell counts as Admitted on each. Self-hosted acceptance ran on Linux machines; the self-hosted guide lists the supported platforms. A verified operation is verified on its own, not in every combination with other options; combinations that a profile rejects are listed in the operation's contract.

Execution and input

Operation Codex Claude Code MiniMax Code
Text Turns, active input, cancellation, restart and continuation Verified on all placements Verified on all placements Verified on all placements
Files and Artifacts Verified: hosted, self-hosted Verified: hosted, self-hosted Verified: hosted, self-hosted
Agent instructions Admitted Admitted At most 32 KiB; longer instructions are rejected
Whitespace-only message text Admitted; delivered unchanged Admitted; native drops it, and an all-blank message becomes (no content) Rejected
Message images PNG and JPEG data URLs: verified on all placements; other data URLs: admitted; remote URLs: rejected PNG and JPEG data URLs: verified on all placements; remote URLs and GIF and WebP data URLs: admitted Rejected
reasoning.effort none, minimal, low, medium, high, xhigh, max: admitted low, medium, high, xhigh, max: admitted none, minimal, low, medium, high, xhigh, max: admitted
reasoning.summary concise, detailed, auto: admitted auto: admitted Rejected
service_tier other than auto default, flex, priority, fast: admitted; ultrafast rejected Rejected Rejected
text.verbosity other than medium Admitted; the native model decides Rejected Rejected
Session update of agent.model, reasoning.effort (including null), service_tier and review: auto_review Applied on the next root Turn; reasoning.effort cannot be reset to null Applied on the next root Turn, except service_tier, which is rejected Rejected
Session update of agent.tools Rejected Rejected Rejected
compaction.async: true, experimental.time_awareness: true and the automations and computer_use tools Rejected Rejected Rejected
Public token usage Measured counters Null Null

Native model parameters and provider protocols per Harness are in model execution.

Tools

Operation Codex Claude Code MiniMax Code
Public functions with text results Verified: none, hosted; admitted: self-hosted Verified on all placements; object-root schemas only Verified: none; admitted: hosted, self-hosted; object-root schemas only; a result must arrive within 120 s (difference)
Function results with images Verified: none, hosted; admitted: self-hosted; remote URLs: rejected Successful results only; PNG and JPEG data URLs: verified on all placements; GIF and WebP: admitted Rejected
Structured output Admitted on all placements Verified on all placements Rejected
Deferred function discovery Rejected Verified: none, self-hosted; admitted: hosted Rejected
Disabled web search and programmatic tool calling Verified: none; admitted: hosted, self-hosted Verified: none; admitted: hosted, self-hosted Verified: none; admitted: hosted, self-hosted
Web search cached or live and programmatic tool calling enabled Admitted on all placements; enabled is rejected together with multi_agent live admitted on all placements; cached and enabled rejected Rejected
Service-origin HTTP MCP Verified: none; rejected elsewhere Verified: none; rejected elsewhere Verified: none, with allowed_tools null and required false only; rejected elsewhere
Environment-origin HTTP MCP Verified: hosted, self-hosted Verified: hosted, self-hosted Verified: hosted, self-hosted; allowed_tools null and required false only
Stdio MCP Admitted: hosted, self-hosted Admitted: hosted, self-hosted; cwd must be the workspace directory and allowed_tools null Rejected
HTTP MCP bearer credentials Verified Verified Verified
Required MCP initialization Verified Verified Rejected
Subagents Verified: hosted; admitted: none, self-hosted Verified: hosted; admitted: none, self-hosted Verified: hosted; admitted: none, self-hosted
Subagents together with functions or HTTP MCP Verified: none; admitted: hosted, self-hosted Verified: functions on none; admitted: HTTP MCP, hosted, self-hosted Admitted; children get none of the Session's functions or MCP servers

Claude structured output requires medium verbosity and combines with multi_agent, Skills, Plugins, capability directories, MCP and tool_search. Claude deferred discovery combines with MCP, Skills, Plugins, capability directories and structured output. Claude Subagents inherit the parent's tools, including deferred functions. Claude Code decides whether tool_search takes effect for the configured model, provider and beta settings. Claude on a workspace placement needs the packaged bridge features for each operation (workspace_functions, workspace_structured_output, workspace_tool_search, workspace_mcp_http).

Environment preparation

These operations need a workspace, so they apply to hosted and self-hosted placements only.

Operation Codex Claude Code MiniMax Code
Initial files, setup commands, Skills and Plugins Verified: hosted, self-hosted Verified: self-hosted; admitted: hosted Verified: self-hosted; admitted: hosted
Capability directories Admitted Admitted Admitted
npm and Python packages Admitted Admitted Admitted
packages.system other than null or [] Rejected Rejected Rejected
Network disabled or restricted Admitted: hosted; rejected: self-hosted Admitted: hosted; rejected: self-hosted Admitted: hosted; rejected: self-hosted
Plugin MCP over stdio Verified: hosted, self-hosted Verified: self-hosted; admitted: hosted Verified: self-hosted; admitted: hosted
Plugin MCP over HTTP Admitted, with literal headers or HTTPS bearer Admitted, with literal headers or HTTPS bearer Verified: self-hosted; admitted: hosted; literal headers or HTTPS bearer

External workspace storage additionally requires the selected Harness declaration and Runtime to support retained_native_history; E2B and OpenSandbox, which keep their own disks, do not. microsandbox always uses external storage, so MiniMax Code, which does not declare the capability, cannot run on it. The Workspace filesystem protocol owns storage combination admission.