Skills issues on Linux
Issues the maintainers labelled both platform:linux and area:skills.
108 issues · 21 open · 87 resolved (81%) · first seen Nov 4, 2025
Is this getting better or worse?
This class of problem is converging. 38 new reports in the last 90 days vs 67 in the 90 before — -43%. The open backlog peaked at 25 in 2026-04 and sits at 21 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
84 of these reports name the Claude Code build they were running, spanning 59 releases. Heaviest reporters:
- Claude Code v2.1.87
- Claude Code v2.1.126
- Claude Code v2.1.81
- Claude Code v2.1.247
- Claude Code v2.1.228
- Claude Code v2.1.215
- Claude Code v2.1.181
- Claude Code v2.1.152
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 32 days across 87 closures. Of the 87 closures with a recorded reason, 25% were closed as completed and 65 as not-planned or duplicate. The most recent completed fix landed in 2026-08.
Workarounds reported by the community
26 of these issues post a workaround someone says worked. The highest-engagement ones:
- /remote-control (/rc) not recognized as built-in command in existing session
- …, kernel 6.8.0-101-generic) - Authentication: logged in via Workaround Starting a new session with from the terminal works correctly.
Found in the description of #28322 · resolved - [BUG] v2.1.78: Protected directory prompt in bypassPermissions has no override — forces hacky workarounds
- …t on every skill creation. 2. No override flag forces worse workarounds Since there's no setting or similar, users who need this workflow (and there are many — skills are a first-class Claude Code feature) will build wor…
Found in the description of #35646 · resolved - Any way to let Claude run /verify and /code-review on its own again?
- …s skill in automated workflows. I am already implementing a workaround for this but +1, bad idea.
Found in the thread of #79282 · still open - [Cowork] User-created skills stored in ephemeral session dirs — silently deleted on cleanup
- …t and should be protected from accidental deletion. Current Workaround As an interim measure, the following workaround is functional: - Created as a centralized, persistent skill store - Symlinked user skills from into C…
Found in the description of #31422 · still open - [BUG] 'claude-api' is consuming entire context
- …f implies), but the permission is a clean per-skill opt-out in the meantime.
Found in the thread of #70062 · 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 1–100
Ranked by community engagement (reactions weighted over comments).
- /remote-control (/rc) not recognized as built-in command in existing session
- [BUG] v2.1.78: Protected directory prompt in bypassPermissions has no override — forces hacky workarounds
- [Bug] Skill plugin scripts fail on first execution with relative path resolution
- /buddy slash command no longer available — returns 'Unknown skill: buddy'
- Any way to let Claude run /verify and /code-review on its own again?
- [BUG] Custom command in ~/.claude/commands/ not recognized
- [BUG] Auto mode injects system reminder that suppresses AskUserQuestion in skills
- [Cowork] User-created skills stored in ephemeral session dirs — silently deleted on cleanup
- [Bug] "workflow" keyword conflicts with tool trigger syntax
- [BUG] 'claude-api' is consuming entire context
- /reload-plugins doesn't rebuild slash-command index
- [FEATURE] design-sync is React-only: add support for Svelte / non-React design systems
- [BUG] /deep-research — default workflow frequently hits "API Error: Server is temporarily limiting requests"
- [Feature Request] Add `disabledSlashCommands` setting to selectively suppress built-in slash commands
- Skills: no escape syntax for literal `$<digit>` in command bodies
- [Bug] Anthropic API Error with skill model switch
- [BUG] allowed-tools in skill SKILL.md only auto-approves the first Bash call per session, subsequent calls prompt for permission
- [BUG] /code-review --comment silently degrades to terminal printing when the forked agent lacks t he posting tools (repro on 2.1.237, after #84093 closed)
- Skill invocation produces hallucinated output instead of following skill instructions
- [BUG] /design-sync slash command not available in Claude Code v2.1.195
- Opus 4.7 tool-use compliance cliff between 24th-26th May — instrumented evidence
- [Bug] claude-api skill consumes entire context window on Haiku model
- [Bug] Skill listing context budget exceeded with truncated descriptions
- Skill subprocess inherits 1M context flag but switches model, fails with 'Extra usage is required for 1M context'
- Skill tool feedback loop: getPromptForCommand body re-injected as isMeta after every model-invoked Skill call
- [BUG] `.claude/skills/**` Edit hangs indefinitely in headless `--dangerously-skip-permissions` sessions (persists across v2.1.114, v2.1.116, v2.1.117)
- Rate limit error on MAX+ plan when invoking slash commands with low usage
- [BUG] path.relative() computed from bash.cwd instead of CLI startup directory causing edition of Plans and local files to fail
- [BUG] skill description frontmatter not surfaced for ~50% of user-defined skills in system-reminder
- [Bug] Skills with same name from different sources not properly namespaced in slash commands
- [Bug] AWK variable syntax ($2) stripped from skill definitions
- [BUG] Claude Code uses inappropriate exclamations "Damn" when missed a needed usage of relevant skill
- [BUG] Issue: /ultrareview returned empty findings 3 times in a row, all 3 free reviews consumed without usable output.
- [BUG] "~/.claude/skills/ YAML files not loaded (v2.1.137 on WSL/Linux)"
- /help lists /tdd ~20 times despite only 2 source files on disk
- [Bug] Loop command repeats identical responses instead of progressing to next steps
- [Feature Request] UltraReview: Add markdown export and cross-session/cross-device findings retrieval
- [BUG] `/ultrareview` fails in bare-repo + worktree container directory
- [BUG] consumed 2 of 3 free ultrareview runs, now i"m told i will be charged
- [BUG] Embedded commands (!`cmd`) in skill/command markdown silently fail in CI environment (claude-code-action)
- [BUG] /ultrareview fails "Could not find merge-base with main" despite local main branch existing with valid merge-base
- [BUG] Plugin skill short-name resolution fails for some plugins ("Unknown skill" error)
- [Bug] Skill resolution prioritizes user-level over project-level skills with same name
- Custom skills in three-level nested directories not discovered
- [BUG] AskUserQuestion silently skipped when explicitly listed in skill's allowed-tools
- [MODEL] Prose hard-wrapping and max-width capping recur across sessions despite CLAUDE.md and saved memory
- Custom skills no longer working after new installation caused a name collision between bundled skill and custom skill
- Plugin skills loaded twice: from marketplace clone and installed cache
- Slash-command skills with !`...` frontmatter expansions fail to load when sandbox cannot create network namespace
- /ultrareview returns empty findings paired with rate-limit error across 3 consecutive runs
- [BUG] /ultrareview failed TWICE
- [BUG] Ultrareview Stopped
- ultrareview: free run consumed when server-side rate-limited (run crashed before producing findings)
- [BUG] Skills with `disable-model-invocation: true` intermittently fail on user invocation in VS Code extension
- Skill Tool Freezes with minimax m2.7 via Ollama
- [BUG] Claude Code cant access skills created and saved in webui
- [BUG] Prompt Injection in "/insights" output
- [BUG] Subagents spawn as team mates
- [BUG] When @-mentioning a file in skill argument, "Read file (NNN lines)" is printed, but file is not in the context.
- [BUG] Custom skill appears twice in available skills list despite single SKILL.md file
- [BUG] Stacked slash-skill invocations load only the first skill on the SDK / --print path
- [BUG] /code-review ignores requested low/medium effort argument, always runs/renders xhigh workflow tier
- [Bug][harness] Read-only SSH enumeration of LXC containers and storage on a provisioned host wrongly blocked
- [BUG] `/ultrareview` crashes before producing findings — 2 consecutive failures, same branch
- [BUG] All official plugin slash commands (/) unavailable — plugin.json missing commands/skills fields
- [Bug] Claude ignores context and provides incorrect feedback on skill implementation
- [BUG] Plugin skill count issue
- [Bug] Anthropic API Error: Rate Limited - Credit Consumed Without Review Generated
- [Bug] /ultrareview crashes on large diffs without producing findings or billable output
- [BUG] `!` preprocessing in slash-command bodies fails in TUI but works in `claude -p` (with `model:` + `context: fork`)
- [BUG] /review deprecation issue:
- [Bug] Quota consumed on /ultrareview crash before findings generation
- [Bug] Ultrareview session crash: "Review crashed before producing findings"
- [BUG] /ultrareview crashes with no findings, consumes free review credit
- /ultrareview consumes free run on Find-phase timeout failure for oversized diffs
- /ultrareview crashes silently — opaque "Review crashed" error with no root cause or mitigation
- [BUG]
- Auto-compact re-injects prior skill invocations with new user message as Input
- Model repeatedly checks wrong system for 'cron jobs' query, becomes argumentative when corrected
- [BUG] Inline plugin skill content not delivered after /clear — "Successfully loaded skill" but no instructions
- Skill tool fails with 'Unknown skill' in dontAsk permission mode
- /skills dialog shows 'No skills found' after moving skill files; requires full restart to fix
- 2.1.89 regression: local slash command autocomplete broken in projects
- [BUG] Project skills not loading in version 2.1.88
- [BUG] Misleading error message when command frontmatter specifies invalid model
- [BUG] /ralph-loop seems totally broken
- [BUG] Project-skill registry stays rooted at the launch checkout after EnterWorktree — a committed SKILL.md in the entered worktree is permanently "Unknown skill" (2.1.241)
- [BUG] Inline !`cmd` preprocessor executes example commands inside fenced markdown code blocks (repro: /plugin-dev:command-development runs npm test)
- [BUG] Skill with disable-model-invocation: true cannot be invoked by user via slash command
- [BUG] slash-command allowed-tools grants are lost when the run resumes from a task-notification
- [BUG] A command file named skill.md in .claude/commands/ silently swallows all top-level commands in that directory (matched as SKILL.md; Windows case-insensitive)
- Bug: /insights narrative analysis fails silently on large...
- [Bug] Anthropic API Error: Incorrect Safety Flag for Non-Security Skill Review
- Claude repeatedly ignored explicit user directives and substituted its own judgment — session 2026-06-28
- /insights output leaks literal <message> tags into rendered response
- [Feature Request] Implement performance monitoring and SLA guarantees for Claude Max subscription tier
- [BUG] Agent skills: frontmatter field does not inject skill content at startup
- [Bug] Custom skills not recognized in .claude/skills directory
- [Bug] Inaccurate skill documentation: find misclassified as unconditionally auto-allowed
- Plugin harness resolves skill base directory to marketplaces/ instead of cache/ path