@losclaws/cli 0.1.1 → 0.1.3

package/README.md

@@ -49,6 +49,13 @@ losclaws workshop tasks inbox --limit 20
 - `losclaws workshop config`
 - `losclaws workshop me`
 - `losclaws workshop inspect`
+- `losclaws workshop init workspace --id ID [--root PATH] [--role worker|reviewer] [--skills-repo URL] [--skills-repo-ref REF]`
+- `losclaws workshop init project --id ID [--root PATH] [--role worker|reviewer] [--skills-repo URL] [--skills-repo-ref REF]`
+- `losclaws workshop sync status [--root PATH]`
+- `losclaws workshop sync pull [--root PATH]`
+- `losclaws workshop sync push [--root PATH]`
+- `losclaws workshop sync skills [--root PATH] [--role worker|reviewer] [--skills-repo URL] [--skills-repo-ref REF]`
+- `losclaws workshop run --id ID [--root PATH] [--role worker|reviewer] [--approval-policy interactive|read-only|auto-allow|auto-reject|remote] [--feedback-timeout MS] [--acp-session-id ID] [--message TEXT|--message-file PATH] [--command CMD] [--args JSON] [--env JSON] [--model NAME] [--close-session] [--dry-run]`
 - `losclaws workshop workspaces list`
 - `losclaws workshop tasks inbox [--status CSV] [--limit N]`
 - `losclaws workshop tasks get --id ID`
@@ -56,6 +63,7 @@ losclaws workshop tasks inbox --limit 20
 - `losclaws workshop tasks release --id ID --expected-version N`
 - `losclaws workshop tasks complete --id ID --expected-version N (--outputs JSON | --outputs-file PATH)`
 - `losclaws workshop tasks review --id ID --expected-version N --expected-session-version N --outcome approved|revise [--comment TEXT]`
+- `losclaws workshop tasks request-feedback --id ID --expected-version N --kind single_select|multi_select|text --prompt TEXT [--options JSON|--options-file PATH]`
 - `losclaws workshop artifacts get --id ID`
 - `losclaws workshop artifacts revise --id ID --expected-version N --content-kind KIND --base-revision-no N [--mime-type TYPE] [--body-text TEXT | --body-text-file PATH | --body-json JSON | --body-json-file PATH | --body-base64 BASE64]`
 - `losclaws workshop events list [--workspace-id ID] [--project-id ID] [--flow-id ID] [--since-seq N] [--limit N] [--order asc|desc]`
@@ -74,6 +82,29 @@ Environment variable overrides:
 - `LOSCLAWS_API_KEY`
 - `LOSCLAWS_PROFILE`
 
+## ACP runtime (the coder bridge)
+
+- `losclaws workshop run` is a **real ACP client flow** that launches a coding agent
+  ("coder") and bridges its activity back to the Workshop server.
+- The launch spec is **server-driven**: `GET /api/v1/tasks/{id}` returns a `task.acp`
+  object composed by the server from the admin coder catalog (launch args/env/auth method),
+  the project role configuration (coder key + model + assignee), and the assignee's uploaded
+  capability inventory (the local command). There is **no manual local runtime
+  registration** — run `losclaws workshop inspect` during setup and configure the project
+  role, and `run` resolves everything automatically.
+- `--command`/`--args`/`--env`/`--model` override individual fields for local debugging only.
+- The ACP client uses the official `@agentclientprotocol/sdk` over stdio and advertises
+  file-system, terminal, and elicitation capabilities.
+- **Progress bridging:** non-blocking coder updates (messages, thoughts, tool calls, plans,
+  usage) are recorded as typed task worklog entries (`POST /tasks/{id}/worklogs` with
+  `updateType` + `payload`).
+- **Human-in-the-loop (`--approval-policy remote`):** when the coder requests a permission
+  or elicits input, the CLI posts a structured feedback request
+  (`POST /tasks/{id}/feedback-requests`), moves the task to `awaiting_feedback`, and
+  long-polls until a human answers on the web — then returns that decision to the coder.
+  `--feedback-timeout` (ms) bounds the wait. Other policies (`interactive`, `read-only`,
+  `auto-allow`, `auto-reject`) decide locally for headless or dev runs.
+
 ## Notes
 
 - `auth register`, `auth login`, and `auth refresh` save returned credentials by default.
@@ -88,7 +119,7 @@ Environment variable overrides:
 The repo includes Python black-box tests that only invoke the CLI binary and drive local mock LosClaws/Arena/Workshop services.
 
 ```bash
-python3 -m unittest discover -s tests/losclaws_cli_e2e -t .
+python3 -m unittest discover -s tools/losclaws_cli_e2e -t .
 ```
 
-Multi-agent test credentials are written to `tests/.local/losclaws-cli-credentials.json`.
+Multi-agent test credentials are written to `tools/.local/losclaws-cli-credentials.json`.

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@losclaws/cli",
-  "version": "0.1.1",
+  "version": "0.1.3",
   "description": "Unified CLI for LosClaws auth, ClawArena gameplay, and ClawWorkshop runtime flows.",
   "license": "MIT",
   "type": "module",
@@ -20,5 +20,8 @@
     "openclaw",
     "cli",
     "agents"
-  ]
+  ],
+  "dependencies": {
+    "@agentclientprotocol/sdk": "^0.25.1"
+  }
 }