Agents issues on Linux — page 3
Issues the maintainers labelled both platform:linux and area:agents.
356 issues · 126 open · 230 resolved (65%) · first seen Feb 7, 2026
Is this getting better or worse?
This class of problem is holding steady. 158 new reports in the last 90 days vs 136 in the 90 before — +16%. The open backlog peaked at 126 in 2026-10 and sits at 126 today.
Derived by counting each matching issue from its creation date until it closed. First-party
computation over the anthropics/claude-code tracker — the trend is not published
anywhere upstream.
Which builds report this
300 of these reports name the Claude Code build they were running, spanning 134 releases. Heaviest reporters:
- Claude Code v2.1.119
- Claude Code v2.1.281
- Claude Code v2.1.234
- Claude Code v2.1.198
- Claude Code v2.1.251
- Claude Code v2.1.260
- Claude Code v2.1.233
- Claude Code v2.1.296
Report volume, not a defect count: an issue can mention a build without that build having caused the failure. Versions come from the version string in the bug-report template, matched against real published tags.
Do these get fixed?
Median time to close is 35 days across 230 closures. Of the 230 closures with a recorded reason, 16% were closed as completed and 193 as not-planned or duplicate. The most recent completed fix landed in 2026-09.
Workarounds reported by the community
106 of these issues post a workaround someone says worked. The highest-engagement ones:
- [BUG] Subagent Write tool rejects .md files named "report"/"summary"/"findings"/"analysis" — no opt-out
- …all users — including those with legitimate workflows — to work around it. The fix could be as simple as a key: . 2. The fallback degrades the multi-agent pipeline. When the Write tool is blocked, the subagent falls bac…
Found in the thread of #44657 · still open - PermissionRequest hooks not triggered for subagent permission requests in Agent Teams
- …The current behavior makes these two features incompatible. Workaround Pre-approve common subagent commands in permissions so they don't need to ask: This lets subagents work autonomously but removes the ability to revie…
Found in the description of #23983 · still open - [BUG] Agent tool isolation: "worktree" has no effect for team agents — agent runs in main repo
- …rktree so the agent works on an isolated copy of the repo." As a workaround, we manually create git worktrees before spawning agents: ...and set in the agent prompt to the worktree path. This works correctly, but the par…
Found in the description of #33045 · resolved - 2.1.178: native team-management tools TeamCreate / TeamDelete no longer surfaced (regression from 2.1.177)
- …in/claude), Linux 6.17.0-14-generic, tool search enabled. - Workaround in place: pinned to 2.1.177 and moved the 2.1.178 binary out of the version pool so the launcher can't re-select it (editing the symlink alone wasn't…
Found in the description of #68721 · resolved - advisor() tool inflates reported input tokens by forwarding full transcript, triggering premature auto-compaction on extended context models
- …ction and destroys the context it was supposed to help with Workaround Setting and/or using a PreCompact hook that blocks auto-compaction: This blocks all auto-compaction and relies on manual . Not ideal but prevents the…
Found in the description of #53065 · still open
Excerpts are quoted from the linked issue so you can judge relevance before opening it. Threads that only ask for a workaround, or report that there is none, are excluded. These are community suggestions, not official guidance from Anthropic — read the full thread before applying one.
Matching issues 201–300
Ranked by community engagement (reactions weighted over comments). Listing the top 300 of 356 matches; the statistics above cover all 356.
- [BUG] Agent-teams — spawned teammates ignore the configured model/effort (settings.json + the lead's model), booting at defaults
- Agent tool `isolation: "worktree"` is silently ignored when spawning teammates with `team_name`
- [BUG] Auto-compaction in a team-lead background session silently switched the main loop to teammateDefaultModel (durable until manual /model)
- [Bug] Agent workflow resumes rerun completed tasks instead of resuming from checkpoint
- Workflow agent count in /workflows UI is off by one
- [BUG] Agent skills: frontmatter field does not inject skill content at startup
- [BUG] Monitor usage creates fake Subagent completion events
- [Bug] Session cost significantly higher than expected; parallel agent forking causing excessive API usage
- [BUG] Custom statusLine permanently stops being invoked after a subagent/response completes at project scope — but not at user scope
- [BUG] Workflow tool determinism guard false-positives on new Date(/Date.now(/Math.random( substrings inside prompt strings
- [Feature Request] Add merge verification option to ExitWorktree for safe worktree cleanup without data loss
- tools: frontmatter — Read(/path), Write(/path), Glob(/path), Bash(cmd:*) parentheticals silently discarded (no runtime enforcement)
- [Bug] Child agent inherits incorrect model version from parent configuration
- [Bug] Agent tool worktree isolation fails with concurrent agents using stale commits
- Garbled/overlapping text in transcript during rapid bash execution with active subagents
- [FEATURE] Surface short session IDs — shell `claude agents list` + show IDs in agent view
- [BUG] Team-inbox `read` flag is set when message is queued, not when agent has actually seen it
- Agent teams: tmux socket unlinked while server alive — ghost servers and leaked tmux-spawn scopes
- [BUG] Team mode subagents don't load system prompts from plugin-defined agents
- [Bug] Quota limit exceeded causes loss of agent work
- Agent Teams: model parameter does not support version pinning
- [Bug] Agent Teams: Sub-agent spawning failures and shutdown handling issues
- [Bug] Tmux teammates not inheriting API key environment variables from parent session
- Sub-agent Read/Edit denied on own worktree path despite bypassPermissions (v2.1.97)
- [BUG] using ultraplan - the remote plan mode container crashes, the agent should either halt and ask, not improvise.
- [BUG] Custom statusline with session name displaces 'View Teammates' tmux command for agent teams
- Agent teams: allow routing plan approvals and permission prompts to user instead of team lead
- Session resume corrupts JSONL: file-history-snapshot messageId collides with message uuid
- [BUG] Vertex/Bedrock subagents silently downgraded to older models (not fully fixed)
- [Bug] Worktree changes lost when agent instances terminate without committing
- [BUG] Agent teammate pane creation fails with "size invalid" on headless server via SSH+tmux
- [BUG] Task Tool File Creation Failure
- [BUG] Built-in subagents (Explore, claude-code-guide, statusline-setup) use hardcoded model names, ignoring parent model
- Agent Teams ignores ANTHROPIC_DEFAULT_OPUS_MODEL, hardcodes members to claude-opus-4-6
- [Bug] Subagent/swarm completion wait hangs after all tasks complete and shutdown
- [BUG] Subagent teardown kills the parent session's stdio MCP server when both declare the same inline server
- [BUG] Subagent transcripts missing final assistant entry (stop_reason / final usage) since 2.1.280, undercounting output tokens
- [Bug] /agents command removed: can no longer create, edit, or delete agents
- Custom subagent frontmatter model (haiku/sonnet) ignored — 1,149 of 1,371 subagents ran on parent Opus model (2.1.277–2.1.290, intermittent)
- [Bug] Opus generates repetitive agent loops and produces faulty tests requiring multiple edits
- Worktree isolation in a background session blocks subagents' Bash commands, even for other repos
- Commands that hit the Bash timeout keep running in the background unbounded; agents re-run them and processes pile up until the host runs out of memory
- [BUG] Agent .md without `name:` in frontmatter is silently skipped: no warning, "Agent type not found" is the only signal
- [Bug] Agent workflow performance degradation on trivial tasks
- [Bug] Subagents fail to terminate reliably
- VS Code extension: Read/Write/Edit fail with "The user doesn't want to take this action right now" after the extension's control stream closes (no user action)
- [BUG] Second SubagentHandback after resume is rejected and the subagent stops instead of reporting via SendMessage — follow-up report is silently lost
- [Bug] Agent spawning restriction not enforced, causing unauthorized agent fork and excessive API usage
- Agent tool honours a `name` parameter that the schema shown to the model omits
- [BUG] Workflow dispatched 334 subagents after a session-limit error; all failed, 111 findings were silently recorded as "refuted", and the run still reported completed
- [BUG] Desktop (Linux): spawn_task sessions still start without claude_desktop_config.json MCP servers; attach happens on UI focus, tools reach the model only on the next queued turn (re: #67432)
- You've hit your session limit · resets 8:10pm (Europe/Rome)
- [Bug] Terminal UI displays incorrect model name when subagents launch or switch models
- [Bug] Unauthorized branch and worktree creation during operations
- [BUG] Four shipped plugin agents fail YAML frontmatter parsing and load as placeholders or not at all
- [Feature Request] Agent Foundry: Build custom agents with framework templates and skill library
- Session renamed in the mobile app never reaches the host, so it cannot be found or addressed by that name
- Session renamed in the mobile app never reaches the host, so it cannot be found or addressed by that name
- [Bug] Subagent retry loop causes excessive context accumulation in ultracode mode
- [Bug] Subagents in ultracode mode ignore window size limits
- [BUG] claude-fable-5 (xhigh): parent session permanently wedges — parallel subagent results never reconcile, event stream goes silent (repro 3x, 0.3.197 & 0.3.251)
- [BUG] --resume sessions register the agent-team hub under a stale placeholder session id — teammate replies are consumed but never delivered
- [BUG] SendMessage-resumed subagent's completion notification routes to the session root, not to the resumer
- SendMessage racing an agent's task stop strands the reply with no notification
- [Bug] Exit cleanup recursively deletes $HOME instead of session directory
- [Bug] Subagent status line JSON missing effort field for non-custom subagents
- [Bug] Subagent status line JSON missing parent reference for nested subagent hierarchy
- Claude Code repeatedly fails to apply its own persisted rules, and allocates effort away from the user's stated goal
- [Bug] Incorrect flagging behavior in Fable 5 integration
- Subagent (Agent tool) result flagged by harness as instruction poisoning: fabricated docs example steering toward .env exfiltration
- [BUG] PR Review skill burns through usage limits in 10 minutes and then loses all progress
- [Bug] Session limit triggered by unexpected agent behavior outside user instructions
- Opus orchestrator contaminates "independent" subagent consults: injects its own candidate answers, relabels an unverified subagent claim as VERIFIED, and transmits a rejected design's frame while warning against it
- [Bug] Subagent final assistant messages not delivered to parent session
- Prohibited block forbids all deletion, unscoped — blocks rm, log scrubbing, and cleanup in coding/ops agents
- Background subagent task-notifications dropped when child completes while parent agent is mid-turn → permanent silent stall of fan-out skills (e.g. /code-review)
- [BUG] Agent teams: in-process teammate window header shows the lead's model, not the teammate's actual model
- Inter-agent wake signals are lossy (edge-triggered, no ack, no idle re-poll) — dispatchers idle forever on completed work
- [BUG] Subagent card stays 'running' after completion notification is delivered; only manual stop freezes it, then persists as a ghost card visible from another session
- [BUG] Parked subagent tool calls time out after exactly 600s and are reported as "The user doesn't want to take this action right now"
- SendMessage: inter-agent messages arrive with no sender provenance, making legitimate coordinator relays indistinguishable from injection
- [FEATURE] Support disable-model-invocation in agent frontmatter (plugin agents load descriptions unconditionally)
- Workflow journal.jsonl result lines carry no agent label — post-hoc joining of results to workflow stages is impossible
- Subagent identity confusion: clicking an agent chip silently redirects user input to the subagent, whose env is byte-identical to the parent (same CLAUDE_CODE_SESSION_ID) — enabled unintended pushes under the parent session's identity
- [BUG] Task store splits after crash + --resume: idle dispatcher replays stale tasks to subagents as synthetic "task-list" messages
- Explore subagent emitted a canned CI-triage prompt template instead of performing its task (0 tool calls)
- All agents restarted because one failed due to reaching the session limit.
- [Bug] Anthropic API: Claude Opus 4.8 Regurgitating Training Instructions in First Turn
- Subagent SendMessage triggers approval popup despite bypassPermissions defaultMode
- [Bug] Agent tool calls silently dropped in parallel <invoke> blocks
- [BUG] Possible bypass of Fable 5 guardrails
- You've hit your session limit · resets 2:20pm (America/Winnipeg)
- Workflow: transient "Connection closed mid-response" in one agent aborts the whole run and discards all completed work
- [BUG] Subagent permission prompts are invisible in remote/web sessions
- [BUG] Headless (-p / stream-json): mid-turn task notifications and hook attachments never reach the model
- [Bug] Long-running agent sessions re-read unbounded context, causing excessive token usage
- [Bug] Auto mode runs shell commands without displaying them
- [BUG] An agent started by $.agent.spawn from a $.clock callback can’t reach its own plugin’s tool.call hook
- [BUG] Registry sweep deletes other hosts' live session records when hosts share $HOME and /etc/machine-id (HPC clusters, cloned images)
- [BUG] ${CLAUDE_EFFORT} in a skill preloaded into a subagent expands to the session effort, not the subagent's effort