feat(automation): add configurable continuation and coordinator access (#2943)

This commit is contained in:
Carlos Florêncio
2026-08-24 22:04:54 +00:00
committed by GitHub
parent 7f7a062a03
commit 29289520c7
126 changed files with 8796 additions and 1790 deletions
+14 -13
View File
@@ -129,19 +129,19 @@ That parser is not a shell and is not quote-aware: quotes and backslashes do not
Select an agent, create a profile, then open **Settings > Agents > _Agent_ > _Profile_**. The page shows the resolved command preview and only the settings supported by that agent.
| Setting | Runtime behavior |
|---|---|
| Name | Label shown in workflow, session, and automation selectors. |
| Model | Requested through ACP when the agent supports model selection. Leaving it unset uses the agent's default where the form allows that. |
| Mode | Requested with ACP `session/set_mode`. The choices come from the installed agent. |
| Configuration options | Dynamic ACP values requested with `session/set_config_option`. |
| CLI flags | Enabled entries are tokenized and appended to the ACP launch command. |
| Command prefix | Optional ACP-only launcher argv prepended to the command, for example `greywall --`. |
| Environment | Literal values or references to Kandev secrets, resolved when the process starts. |
| CLI passthrough | Uses the CLI's native terminal interface instead of a structured ACP conversation. |
| Enabled | Keeps the profile available to existing sessions and settings while hiding it from new task, session, handoff, and Quick Chat selectors. |
| Setting | Runtime behavior |
| ---------------------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| Name | Label shown in workflow, session, and automation selectors. |
| Model | Requested through ACP when the agent supports model selection. Leaving it unset uses the agent's default where the form allows that. |
| Mode | Requested with ACP `session/set_mode`. The choices come from the installed agent. |
| Configuration options | Dynamic ACP values requested with `session/set_config_option`. |
| CLI flags | Enabled entries are tokenized and appended to the ACP launch command. |
| Command prefix | Optional ACP-only launcher argv prepended to the command, for example `greywall --`. |
| Environment | Literal values or references to Kandev secrets, resolved when the process starts. |
| CLI passthrough | Uses the CLI's native terminal interface instead of a structured ACP conversation. |
| Enabled | Keeps the profile available to existing sessions and settings while hiding it from new task, session, handoff, and Quick Chat selectors. |
| Auto-approve all permissions | Answers automatically: the first `allow_once`/`allow_always` option, otherwise the first option supplied by the agent; no options cancels. It is off by default. |
| MCP servers | Adds profile-specific external MCP servers when the agent supports MCP. |
| MCP servers | Adds profile-specific external MCP servers when the agent supports MCP. |
Model, mode, command, and configuration choices are probed from the locally installed CLI and cached. The managed **Update agent** action refreshes them automatically; after other CLI changes, refresh the profile manually. Probe status can report **auth required**, **not installed**, **not configured**, or **failed**; a saved model name does not prove that the current provider account can use it.
@@ -201,6 +201,7 @@ candidate is eligible, the session waits for a recovery action. After the
current turn settles, use **Retry current agent** or **Try next agent** in the
session recovery surface. These actions use the current route generation, so a
stale browser action does not replace a newer route decision.
### Host probes and executor model catalogs
The model list shown while editing a profile comes from a host probe. It is an
@@ -318,7 +319,7 @@ environment values, headers, raw MCP arguments, or option metadata. See
Auto approval can authorize shell commands, file changes, network calls, or any other capability exposed by that agent. Agent-specific flags that suppress permission prompts can be broader still. Use either only with a constrained executor, repository, environment, and credential set.
Workspace automation selectors do not offer passthrough agent profiles or Local executor profiles. **Run**-mode automations also cannot wait for a permission response: an unanswered request is rejected and the run fails. Use **Task** mode when a person must approve agent actions, or use a profile whose safe work does not prompt. See [Automation and MCP](automation-and-mcp.md).
Workspace automation selectors do not offer passthrough agent profiles. Local executor profiles are available for the repository-free target; Worktree requires a repository. Hidden automation sessions receive a fixed workspace-scoped coordinator MCP surface. Visible normal-task automations use the ordinary task profile and MCP surface. The trusted automation principal is resolved before hidden-run dispatch, and a hidden automation task and its sessions cannot be used as mutation, messaging, stopping, spawning, or blocker targets. Cross-task spawning uses the target task's normal profile. Native provider continuation and compaction remain authoritative for a healthy reusable session; Kandev's fallback resume prompt uses only the newest 50 non-empty user or assistant messages and excludes tool events. See [Automation and MCP](automation-and-mcp.md).
## Structured ACP and terminal passthrough