Upgrade Cline SDK to v0.0.72 and handle durable queues plus run.failed
Move integrations to Cline SDK v0.0.72 so queued prompts survive aborts, sessions resume with prior context after hub restarts, failed queued turns emit run.failed, and MCP/LiteLLM/network edge cases stop dropping work.Why this loop
v0.0.72 changes abort, queue, and failure contracts that current hosts likely still assume. Prompts queued mid-turn are kept across user aborts, drained after a self-abort, and accept edits made inside the abort window instead of being dropped. Session stop is a consistent full stop across hosts. Session context is durable across aborts and hub restarts, so resume must restore the interrupted state rather than a reset. Failed queued turns now emit run.failed instead of completing silently — any handler that treats queue drain as success will hide errors. A hung MCP server no longer kills session creation; unconfigured stdio MCP gets a 30s initialize budget. Remote SSE MCP surfaces an OAuth prompt on 401 and supports pre-registered OAuth clients. LiteLLM is routed through Chat Completions, not the Responses API. Mid-stream network drops before any model output are retried. Vertex ADC refreshes use the configured fetch. Checkpoint diffs include files that were untracked at snapshot time, and checkpoints are picked up if git is initialized mid-session. Plugin settings/contributions are hub-centralized with host-aware snapshots and atomic host toggles. Scheduled run reports now carry headers, schedule metadata, durations, and lifecycle errors.
Proposed actions
- Pin the Cline SDK to the sdk/sdk/v0.0.72 release (https://github.com/cline/cline/releases/tag/sdk/sdk/v0.0.72), bump the dependency and lockfile in every host/integration that imports the SDK, and restart those processes so they load the new abort-queue, session-durability, and run.failed code.
- Find every queued-turn completion path and change it so a failed queued turn is handled as run.failed: do not treat silent completion or queue drain as success; log, surface, and fail the item only on a terminal run.failed (or explicit success).
- Add/adjust abort tests the agent can run: (1) queue a prompt during an in-flight turn then user-abort — the queued prompt and any queue edits made in the abort window must still be present; (2) after a turn self-aborts, confirm the remaining queue is drained rather than discarded; (3) abort or restart the hub mid-session and assert resume uses the pre-interrupt context, not a reset.
- If you call LiteLLM proxies, assert requests go through Chat Completions and not the Responses API. If you use remote SSE MCP, handle HTTP 401 by showing the OAuth authorization prompt and configure pre-registered OAuth clients where dynamic client registration is unavailable.
- For MCP hosts: a hung MCP server must not fail session creation; unconfigured stdio servers must get a 30-second initialize timeout instead of blocking forever. Retry network interruptions that occur mid-stream before any model output. Confirm Vertex ADC token refresh uses the configured fetch (proxies/custom transports). Verify checkpoint diffs include files that were untracked at snapshot time and that checkpoints appear if git is initialized part-way through a session.
Agent prompt
Paste into your agent or query via MCP (get_agent_prompt) — free, no extra AI costCline / .clinerules task
DevAgentRadar → Cline
You are helping me adopt a real coding-assistant change. Work only from the facts below. Do not invent features.
Context
Assistant: Cline Proposal: Upgrade Cline SDK to v0.0.72 and handle durable queues plus run.failed Summary: Move integrations to Cline SDK v0.0.72 so queued prompts survive aborts, sessions resume with prior context after hub restarts, failed queued turns emit run.failed, and MCP/LiteLLM/network edge cases stop dropping work. Primary source: https://github.com/cline/cline/releases/tag/sdk/sdk/v0.0.72
Why it matters
v0.0.72 changes abort, queue, and failure contracts that current hosts likely still assume. Prompts queued mid-turn are kept across user aborts, drained after a self-abort, and accept edits made inside the abort window instead of being dropped. Session stop is a consistent full stop across hosts. Session context is durable across aborts and hub restarts, so resume must restore the interrupted state rather than a reset. Failed queued turns now emit run.failed instead of completing silently — any handler that treats queue drain as success will hide errors. A hung MCP server no longer kills session creation; unconfigured stdio MCP gets a 30s initialize budget. Remote SSE MCP surfaces an OAuth prompt on 401 and supports pre-registered OAuth clients. LiteLLM is routed through Chat Completions, not the Responses API. Mid-stream network drops before any model output are retried. Vertex ADC refreshes use the configured fetch. Checkpoint diffs include files that were untracked at snapshot time, and checkpoints are picked up if git is initialized mid-session. Plugin settings/contributions are hub-centralized with host-aware snapshots and atomic host toggles. Scheduled run reports now carry headers, schedule metadata, durations, and lifecycle errors.
Suggested actions
- Pin the Cline SDK to the sdk/sdk/v0.0.72 release (https://github.com/cline/cline/releases/tag/sdk/sdk/v0.0.72), bump the dependency and lockfile in every host/integration that imports the SDK, and restart those processes so they load the new abort-queue, session-durability, and run.failed code.
- Find every queued-turn completion path and change it so a failed queued turn is handled as run.failed: do not treat silent completion or queue drain as success; log, surface, and fail the item only on a terminal run.failed (or explicit success).
- Add/adjust abort tests the agent can run: (1) queue a prompt during an in-flight turn then user-abort — the queued prompt and any queue edits made in the abort window must still be present; (2) after a turn self-aborts, confirm the remaining queue is drained rather than discarded; (3) abort or restart the hub mid-session and assert resume uses the pre-interrupt context, not a reset.
- If you call LiteLLM proxies, assert requests go through Chat Completions and not the Responses API. If you use remote SSE MCP, handle HTTP 401 by showing the OAuth authorization prompt and configure pre-registered OAuth clients where dynamic client registration is unavailable.
- For MCP hosts: a hung MCP server must not fail session creation; unconfigured stdio servers must get a 30-second initialize timeout instead of blocking forever. Retry network interruptions that occur mid-stream before any model output. Confirm Vertex ADC token refresh uses the configured fetch (proxies/custom transports). Verify checkpoint diffs include files that were untracked at snapshot time and that checkpoints appear if git is initialized part-way through a session.
After you finish
Do not report this as applied to DevAgentRadar. You cannot write the visitor's loop.
Tell the human: open https://devagentradar.com/proposals/cline-sdk-sdk-v0-0-72-upgrade-cline-sdk-to-v0-0-72-and-handle-durable-qu and mark Applied, Skipped, or Failed. Proposal id: 31a52e73-7e73-4de9-9556-18de8de3e653
Your job
- Restate the change in one sentence.
- Propose a minimal plan for my repo (or a throwaway pilot).
- Implement only what I approve; prefer small diffs and tests.
- Call out risks (permissions, breaking APIs, cost).
Start by confirming you understood the proposal.
Your loop
This browser · no sign-in · not shared as “you”After you run the prompt
Only you can mark this. Agents cannot write your loop.
Your decision stays on this device. A public tally appears after a few votes.