Skills issues on macOS
Issues the maintainers labelled both platform:macos and area:skills.
474 issues · 79 open · 395 resolved (83%) · first seen Feb 23, 2026
Is this getting better or worse?
This class of problem is converging. 161 new reports in the last 90 days vs 289 in the 90 before — -44%. The open backlog peaked at 117 in 2026-04 and sits at 79 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
313 of these reports name the Claude Code build they were running, spanning 114 releases. Heaviest reporters:
- Claude Code v2.1.119
- Claude Code v2.1.87
- Claude Code v2.1.112
- Claude Code v2.1.63
- Claude Code v2.1.207
- Claude Code v2.1.92
- Claude Code v2.1.76
- Claude Code v2.1.233
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 395 closures. Of the 395 closures with a recorded reason, 24% were closed as completed and 299 as not-planned or duplicate. The most recent completed fix landed in 2026-08.
Workarounds reported by the community
112 of these issues post a workaround someone says worked. The highest-engagement ones:
- [Bug] /loop slash command not recognized in v2.1.71
- …d on a recurring interval (e.g. /loop 5m check the deploy)" Workaround The underlying CronCreate tool works correctly and achieves the same functionality: - CronCreate with cron expression /5 and a prompt successfully sc…
Found in the description of #31759 · resolved - AskUserQuestion auto-submits without waiting for user input in skill/command context
- …t was never collected until they review the final artifact. Workaround implemented Added defensive instructions to skill workflow prompts: after each AskUserQuestion call, verify the result contains actual option labels.…
Found in the thread of #29674 · resolved - [BUG] Skill `model:` frontmatter triggers 429 "Extra usage required" on 2.1.76 when session uses opus[1m]
- …Sonnet promoted to . Rejected if account lacks entitlement. Workarounds - Stay on 2.1.75 — skill model switching works correctly without auto-promotion. Caveat: if a sonnet skill (200K) is invoked from an opus[1m] sessio…
Found in the description of #34296 · still open - [BUG] permissions.additionalDirectories does not load skills from added directories (--add-dir CLI flag does)
- …ilable-skills list. Only bundled skills ( , , etc.) appear. Workaround: Restored — all 22 skills load fine via the personal-scope path. Confirming this is identical symptom to the Windows repro in 43267 (auto-closed as d…
Found in the thread of #30064 · resolved - Regression: built-in `/code-review` rejects invocation from another skill (disable-model-invocation), breaking workflow composition
- …estore this without changing the safer default. Our current workaround is to pause the workflow and hand the developer an explicit manual step, then resume with the output folded in — but that reintroduces exactly the ma…
Found in the thread of #79560 · resolved
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). Listing the top 300 of 474 matches; the statistics above cover all 474.
- [BUG] /buddy returns "Unknown skill: buddy"
- [Bug] /loop slash command not recognized in v2.1.71
- AskUserQuestion auto-submits without waiting for user input in skill/command context
- Workflow tool triggers on keyword 'workflow' in normal conversation
- [BUG] Skill `model:` frontmatter triggers 429 "Extra usage required" on 2.1.76 when session uses opus[1m]
- [BUG] permissions.additionalDirectories does not load skills from added directories (--add-dir CLI flag does)
- Regression: built-in `/code-review` rejects invocation from another skill (disable-model-invocation), breaking workflow composition
- AskUserQuestion auto-submits with empty response when called from custom commands/skills
- AskUserQuestion UI not rendered when called after Skill tool invocation
- /code-review skill: silent fallback to main...HEAD reviews other people's commits, and JSON-only output is hard to read
- [Bug] Agent unable to invoke /code-review skill from custom skills
- ultrareview crashed before producing findings — free credit consumed
- AskUserQuestion auto-completes with empty answers when listed in command allowed-tools
- [BUG] Slash commands with Skill() in allowed-tools trigger rate limit on minimal diffs (regression)
- [BUG] Plugin skills not appearing in `/` menu after v2.1.110 update — only `schedule` shows
- [Bug] deep-research workflow aborts entire run (and burns millions of tokens) when any schema-bound subagent fails to emit StructuredOutput
- EnterWorktree tool creates worktrees inside .claude/ directory, breaking slash command and skill discovery
- [BUG] Skills from ~/.claude/skills/ and plugins not shown in Customize -> Skills panel
- Project-level skills in .claude/skills/ are now uneditable — protection conflicts with core feature design
- /ultrareview fails: seed bundle import error from Firestore
- [Bug] Ultrareview quota consumed on server-side rate limit failures
- [BUG] AskUserQuestion auto-selects without waiting for user input in custom commands/skills
- [BUG] Skill with paths frontmatter is completely undiscoverable
- [BUG] 2.1.129 prints "47 skill descriptions dropped" — same files load clean on 2.1.128
- [BUG] Write permission allowlist ignores relative file_path that resolves to absolute-path rule target (still broken in 2.1.122 — see #29098 / #37220 / #38391)
- [BUG] argument-hint frontmatter not displayed for plugin skills in slash command autocomplete
- [BUG] Organisation skills uploaded via admin console do not appear in Claude Code CLI
- claude.ai Skills and Cowork plugins silently injected into Claude Code context — no opt-out, ~6k tokens wasted per session
- AskUserQuestion silently returns empty when called in the same turn as a Skill invocation
- Desktop app Customize window doesn't show user-authored skills from ~/.claude/skills/
- ${CLAUDE_SKILL_DIR} not substituted in SKILL.md frontmatter hook commands
- [FEATURE] Project-scoped skills in Cowork (parity with Claude Code's `.claude/skills/`)
- Global commands from ~/.claude/commands/ not appearing in slash command menu
- [BUG] Skills with `disable-model-invocation: true` cannot be invoked by subagents even when parent agent explicitly references the skill
- Skill hot-reload doesn't detect new skill directories added mid-session
- Desktop app embedded loader ignores plugin manifest 'skills' array — serves all repo skills under every plugin, SHA-namespaced (CLI 2.1.206 correct)
- Skills are siloed across Chat / Cowork / Code tabs in the desktop app
- [BUG] /compact and auto-compaction drop entire Available skills system-reminder; /reload-skills recovers with "no changes"
- [FEATURE] Claude Code sessions cannot publish/update skills to the account Skills store - expose the Chat surface's "Save skill" flow
- [BUG] /ultrareview crashes before producing findings, but consumes quota
- Claude Code repeatedly commits/pushes without user approval despite explicit CLAUDE.md instructions
- AskUserQuestion widget renders only header then hangs in VS Code extension (macOS, called from Skill)
- [BUG] Plugin install no longer copies symlinked skills to cache (regression in v2.1.117)
- [BUG] Cowork only loads 3/27 personal skills from ~/.claude/skills/ — UI does not scan directory
- AskUserQuestion returns empty answer when PreToolUse hook + skill allowed-tools are both present
- [BUG] Mid-message user-typed slash command with disable-model-invocation: true is silently ignored despite UI highlighting
- Opus 4.7 repeatedly ignored explicit user instructions to use installed skills; caused excessive token spend over multi-hour session
- [Feature Request] Restore `/buddy` command
- Regression: SessionStart hook and skills listing re-injected on every --resume (fixed in v2.1.69/v2.1.70, broken again in v2.1.74)
- AskUserQuestion returns empty inside Skill tool (GSD plugin context)
- AskUserQuestion tool silently auto-resolves with empty responses after Skill tool invocations
- Billing: /ultrareview consumed 2 free runs on server-side rate-limit error
- Skills with context: fork still inject all MCP tool definitions into subagent, ignoring allowed-tools
- Regression: .claude/skills/ writes prompt for confirmation in bypassPermissions mode since 2.1.78
- [Bug] Agent context fills unexpectedly when calling `/claude-api` endpoint
- Auto-update removes user symlinks from ~/.claude/skills/
- Plugin skill `context: fork` + `agent:` frontmatter not honored in 2.1.112 — runs inline instead of dispatching to subagent
- [BUG] `matcher: "Skill"` no longer works for hooks
- Desktop app: slash command autocomplete missing user-created commands
- Custom commands in .claude/commands/ not discovered after upgrading to 2.1.88
- skill-creator: description optimization requires ANTHROPIC_API_KEY with no fallback to claude -p
- [Bug] Internal command shadowed by plugin skill with same name
- [Bug] OTel: skill_activated masks skill.name to "custom_skill" while api_request events and cost/token metrics export it verbatim (regression since ~2.1.126)
- [Bug] Manual-only plugin skills not registered at startup without /reload-plugins
- [BUGdeep-research workflow spawned 111 agents for a single query and consumed entire 5-hour session quota — zero output produced
- argument-hint frontmatter in skills not displayed in slash command hint area
- Ultrareview: free credit consumed despite rate-limit failure (no report produced)
- [BUG] /team-onboarding scans only the current workspace's transcripts, not the user's full Claude Code usage
- [BUG] Marketplace plugins fail to load skills in Cowork -- zip upload of same plugin works fine
- Rate limit triggered by skill invocation despite 90% context remaining
- [BUG] running claude with --agent doesn't preload skills like subagents
- [Bug] /loop /clear command not working
- [BUG] Skills set to `context: fork` output getting swallowed up by Desktop UI
- [BUG] /schedule references https://claude.ai/code/routines but the URL 404s
- [Bug] /ultrareview GitHub authentication fails despite valid gh credentials
- Project-level skills in git worktrees not discovered since 2.1.104
- [BUG] Skills always load from repo root instead of worktree's working directory
- /code-review regression since v2.1.235: background forked-skill execution — parent stalls on last angle, and ReportFindings UI is lost
- [BUG] "workflow" keyword in user message forces model to invoke Workflow tool even in purely conversational/investigative context
- [Feature Request] Restore /buddy command functionality
- Skill descriptions silently dropped from system reminder for ~half of a plugin's skills
- [BUG] /ultrareview reports "GitHub App must be installed" despite correct install on the repo
- Subagent `skills:` frontmatter preload does not inject skill content when the skill has `disable-model-invocation: true`
- [BUG] [ultrareview] Remote review returns empty findings ([]) on large-scope repositories
- [BUG] Claude Desktop 1.2773.0 OOM crash on startup — skills sync memory leak
- [BUG] CRITICAL: P0 .skill file "Save and Replace" silently fails. Claude Cowork packages the skill from the host-mounted path (which had the OLD files), not from the VM working copy where the edits actually lived causes cascading scheduled task failures.
- [BUG] What happend to buddy?
- [BUG] CC forgets mission critical steps in commands, then writing code to main, disobeying strict instructions on multiple layers. This is horrible.
- [BUG] slash commands not discovered in v2.1.88
- Disabled plugins' skills still appear in /skills list
- [BUG] Skill loader ignores installPath for string-source marketplace plugin
- disable-model-invocation skills silently unavailable when skills-dashboard GrowthBook flag is off
- skillOverrides silently ignored for plugin-provided skills — runtime hard-codes source==="plugin" to "on" while the /skills UI implies overrides work
- Two first-party `/code-review` commands (bundled skill + official plugin) collide with contradictory docs
- [Bug] Agent ignores user instructions, commands, and skills
- [BUG] Cowork — account enrolled in two A/B variants (`testfoo` + `0526`) causing skill misfires + connector failure; `ENABLE_TOOL_SEARCH` mismatch
- [BUG] Ultrareview failed and returned empty results — requesting refund of 1 free ultrareview
- "N skill descriptions dropped" — empty descriptions for duplicate SKILL.md between plugins/marketplaces/ and plugins/cache/
- Model silently substitutes a different approach than the one specified, without surfacing the deviation
- /ultrareview: free run consumed when cloud session crashes from upstream rate limiting