anthropics / anthropics/claude-code
Cloud routines (CCR) silently fail, stall in PENDING for hours, or fire duplicates with no visible error
Nobody has claimed this yet.
- Dominant language
- TypeScript
- Stars
- 147k
- Forks
- 24k
- PR merge metrics
- PR metrics pending
Description
Summary
A scheduled cloud routine (Claude Code Routines / CCR) intermittently fails to execute at all,
takes multi-day delays to execute, fires duplicate/extra times for a single logical trigger, or
gets permanently stuck in a PENDING run status — all with no error surfaced via the API or UI.
Environment
- Feature: Claude Code Routines (RemoteTrigger API / claude.ai/code/routines)
- Environment: env_01Vy98YL9GbGkSMxnGW4S6R2 (claude-code-default)
Observed failure modes (with repro evidence)
1. Silent no-op — container + Claude Code start, but the initiating prompt never appears to
execute (no tool calls, no output, nothing in Slack/Confluence that the prompt asked for).
Trigger: trig_01DWjLkBzq4fuWmaa3UE6QwQ (minimal test: "DM yourself the word test").
Affected sessions, all zero output: cse_01DNkKb5UhUeYPieHtUyujdd, cse_019wx1NxaWM88YTt2NEcXJvd,
cse_01PbvSENXUUZi3ToFPPUzsvE, cse_016MKuyw5FKZBnU39HhLnH3K, cse_0142knfUUrjYDaKAZC4ZiEAK,
cse_01D8T4UhcR37j65oaYPjhwYr, cse_01CkhpAUVtsYvQ56r3836dp6.
Run history UI for these shows "Completed" with a transcript that stops right after
"Started Claude Code" — no further steps, no error.
2. Multi-day execution delay — a manually triggered run (trig_018XN6mfKGUkV71AYANLDhNZ,
session cse_01TREXaa1Eqp4SEe6AhDPAoF, fired 2026-08-07T19:46:49Z) produced no output until
2026-08-10 ~15:37 EDT — roughly 3 days later — when it finally executed correctly.
3. Unexplained duplicate fires — same trigger fired twice in one day with no second request
from the user or API caller: 2026-08-14 13:55 UTC and 14:59 UTC. Both executions actually ran
and produced (correct) output, resulting in two duplicate Slack DMs to the same recipient.
4. Stuck PENDING — 2026-09-18: the cron fired on schedule (13:12 UTC, session
cse_014rXdxmHhxHvwVcg6zHwHsf) and remained in ROUTINE_RUN_STATUS_PENDING for 7+ hours with
zero output. A manual retrigger (session cse_01UunYfZmdnfV4oe24C6pQKn, 20:16 UTC) also got stuck
in PENDING.
Impact
Routines cannot currently be relied on for time-sensitive, unattended scheduled tasks — every
success we've observed has ultimately required manual intervention (either triggering, or
verifying and re-triggering) rather than working unattended as designed.
Expected behavior
A routine run should either complete with visible output/tool calls, or surface a clear error/
failure status — not remain in an indefinite PENDING state or silently no-op.
🤖 Generated with Claude Code
Contributor guide
No contributing guide indexed for this repository
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Research direction
Start with the RemoteTrigger API and the claude.ai/code/routines run history for the listed trigger and session IDs, comparing silent no-ops, delayed runs, duplicate fires, and ROUTINE_RUN_STATUS_PENDING cases. Done means each run either executes with visible output and tool calls or surfaces a clear failure instead of silently completing or remaining pending indefinitely.
Written by the indexing model from the issue text.
Assessment
- Domain
- api, cloud, distributed-systems
- Issue type
- Bug
- Difficulty
- 5/5
- Estimated time
- Over a week
- Activity status
- Active
- Clarity
- Needs clarification
- Newbie friendliness
- 25/100