Skills issues on Windows — page 2
Issues the maintainers labelled both platform:windows and area:skills.
212 issues · 50 open · 162 resolved (76%) · first seen Jan 5, 2026
Is this getting better or worse?
This class of problem is holding steady. 65 new reports in the last 90 days vs 84 in the 90 before — -23%. The open backlog peaked at 50 in 2026-10 and sits at 50 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
119 of these reports name the Claude Code build they were running, spanning 88 releases. Heaviest reporters:
- Claude Code v2.1.150
- Claude Code v2.1.126
- Claude Code v2.1.92
- Claude Code v2.1.119
- Claude Code v2.1.123
- Claude Code v2.1.108
- Claude Code v2.1.96
- Claude Code v2.1.266
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 33 days across 162 closures. Of the 162 closures with a recorded reason, 15% were closed as completed and 137 as not-planned or duplicate. The most recent completed fix landed in 2026-10.
Workarounds reported by the community
67 of these issues post a workaround someone says worked. The highest-engagement ones:
- [Feature Request] Support the standard .agents/skills/ directory for agent skills
- …ex, Cursor, ...), with no vendor-specific copies. - Current workarounds (sync tools, symlinks into ) duplicate content or rely on hacks, and create extra work: drift, conflicts, duplicate search results. Proposed behavio…
Found in the description of #16345 · still open - [BUG] Cowork /schedule skill fails — create_scheduled_task tool not injected into session context (Windows)
- Preflight Checklist - [x] I have searched [existing issues]( and this hasn't been reported yet - [x] This is a single bug report (please file separate reports for different bugs) -
Found in the description of #29022 · still open - [BUG] /claude-api bundled skill saturates context unconditionally — ~77% spike from neutral question
- …has the right file structure — it just needs a router. ▎ ▎ Workaround: A user command at ~/.claude/commands/claude-api.md overrides the bundled skill and implements on-demand loading. Effective but fragile — requires ma…
Found in the description of #63566 · resolved - Skill tool call returns 'Tool result missing due to internal error' and agent gets stuck
- …aming hangs indefinitely - 26224 — General hanging/freezing Workaround User must Ctrl+C and manually re-invoke the skill. The CLAUDE.md directive to "immediately retry on internal error" is not followed because the agent…
Found in the description of #43866 · resolved - [BUG] Invoking a skill (slash command) pins a large injected block and shrinks the assistant response into a tiny scroll area (VS Code extension)
- …. - So there's no userland (hooks/settings/skill-authoring) workaround — this needs a fix in the extension's message renderer itself: apply the same "Show more" collapse to skill-injected content that already applies to…
Found in the thread of #72590 · 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 101–200
Ranked by community engagement (reactions weighted over comments).
- Bug: anthropic-skills:docx hangs silently on Windows 11 (Write permission denied)
- /telegram:access skill ignores TELEGRAM_STATE_DIR — breaks multi-instance pairing
- Desktop app: "Could not load skill files" for custom marketplace plugins
- [BUG] 작성한 스킬의 description이 스킬목록에서 보이지 않아.
- Custom skills not visible in slash commands or Settings > Skills UI
- [Bug] Built-in `/bug` command intercepted by user-defined skills in remote sessions
- [BUG] `/security-review` command fails in git repository.
- bug(vscode): Skill tool crashes with EEXIST when home dir is used as working directory on Windows
- model: frontmatter in custom commands doesn't manage context for target model's window
- /gsd:progress and file write operations trigger recurring 500 Internal Server Error from Anthropic API
- [BUG] Auto-compaction still drops the entire Available skills listing (2.1.281, Windows) — #74990 was bot-closed, not fixed
- Skill/command argument substitution is off by one and limited to $0/$1
- Leading multi-slash prompts expand inconsistently (31/35 only the first, 4/35 all); UserPromptSubmit hooks can't see which were dispatched
- [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] /code-review skill is now being automatically loaded in sub-agents the code-review skill spawns. Causing it to keep recursively making more agents
- Workflow tool: deep-research synthesis step accepts schema-valid-but-empty (placeholder) agent output as final result
- /deep-research and /workflows commands unusable with official Claude auth, but work via third-party API (e.g. Kimi)
- Assistant message text hidden when a turn ends with ScheduleWakeup
- Comandos com model: haiku no frontmatter não forçam execução em Haiku quando usuário está em outro modelo
- [BUG] `/loop` scheduled cron fires at wrong time — offset from session start instead of declared cron minute
- [BUG] /insights output renders \.claude path incorrectly on Windows due to markdown escape
- user-invocable: false is ignored by the per-plugin browser menu
- ultrareview crashed before producing findings — refund request (Windows, 2.1.161)
- [BUG] [VS Code Extension] Organization-level skills not loaded (only filesystem paths checked)
- [BUG] Slash command autocomplete should prioritize built-in commands over skills with the same prefix
- /insights: prescriptions need repo-grounding; one trust-eroding security false-positive
- [Windows] Skill shell execution opens interactive mintty (Git Bash) window that does not close
- [BUG] Files over some size threshold (in my skill's case, a file >614 lines / >23,458 bytes) get cut off in the Linux-sandbox view, preventing skill from executing on Windows
- "No repos match" when installing GitHub skill despite app connected and MCP workaround attempted
- [ultrareview] silently returns empty findings on large branch diffs (no scope warning, burns free quota)
- [Bug] bash: .claude/hooks/skill-instructions-hook.sh: No such file or directory
- [Feature Request] Document local scope custom skills/plugins installation
- [BUG] Cowork sub-agent permission resolver translates Linux mount paths to non-existent Windows paths
- [BUG]VS Code extension: slash command triggers API Error: Content block not found
- [Bug] Ultrareview quota decremented on API rate limit failure
- [BUG] /ultrareview consumed free credit on timeout without producing any output
- Feature request: auto-enter plan mode for slash commands / sub-agents
- Skills pane shows empty state during cloud-sync, indistinguishable from "no skills configured"
- [BUG] Enabled scheduled tasks no longer surface in available-skills list (Desktop 1.5354.0 / 1.5220.0)
- [BUG] Skill autocomplete no longer filters by substring match (regression)
- C:/Program Files/Git/ultrareview fails with 502 (and previously 404) — bundle upload broken
- [BUG] SSH session startup: 13+ min per-session delay from sequential per-file plugin sync + no UI toggle for bundled Anthropic skills
- Discover project commands/skills from additional working directories (--add-dir)
- [Bug] skillOverrides in user/project settings has no effect — g7H() is a stub returning "on"
- [BUG] Windows desktop app rejects dotted slash commands (e.g. /smithy.fix) — parses "." as a separator
- Model hallucinates inline tool XML with prior-session content when skill Bash permission fails
- Skill name collision across plugins: fully qualified name resolves to wrong plugin
- [Bug] /insights command: Model omits file:// link from "output exactly" message block
- [BUG] .claude/skills/ not exempt from protected directory prompt on Windows (v2.1.94)
- [BUG] skill-creator: scripts fail on Windows when Python is managed by uv
- Marketplace plugin skill gets silently mapped to official plugin skill instead of registering separately
- [BUG] Custom commands disappear from autocomplete after first use — requires CLI restart
- [BUG] --permission-mode bypassPermissions / --dangerously-skip-permissions not working in print mode (-p) with --resume
- document-skills: soffice.py crashes on Windows (AF_UNIX AttributeError)
- [BUG] Skills from `additionalDirectories` (settings.json) are not loaded
- [BUG] Custom slash commands from `.claude/commands/` no longer appear in autocomplete
- [BUG] Skills visible in Customize UI but not invocable as slash commands when located in project directory
- [FEATURE] Desktop: no way to disable anthropic-skills:* skills in Code tab sessions
- [BUG] Uploading skills to Claude Desktop running via gateway only transfers the SKILL.MD file not other files or folders
- [BUG] Mobile app does not show skill arguments in the user message
- [Bug] Failed to install gitrepos skill dependency
- Skill loading: $1-$19 in a skill file are replaced with unrelated conversation text (CLAUDE.md is unaffected)
- [BUG] Desktop and web clients resolve /plan to built-in plan mode; CLI correctly resolves it to the project's /plan command
- [BUG] Settings UI only shows 1 skill despite 56 skills installed and functional
- Skill/slash command names containing a space only match on the first word
- [BUG] Desktop (Windows/WSL): detached window's slash menu lists only bundled skills
- /schedule fails to connect to remote backend; retries return ambiguous non-error/non-success responses
- Skill tool: multi-word args cause $N positional substitution to corrupt literal $N tokens in skill body (e.g. Excel cell refs)
- [BUG] Assistant's turn overruns and writes a fake "user" message that gets treated as real input
- [BUG] claude-api reinvocation duplicates ~350K-token payload; later turns read ~778K and force >1M compaction
- [Bug] Compact limit error triggered despite available context and free space
- Skill-listing context budget: document trim policy + expose a registered/dropped diagnostic + supported suppression flag
- skill-creator: run_eval.py uses select() on a pipe, fails on all Windows runs (WinError 10038) and reports a misleading score
- [BUG] `xlsx` skill's `recalc.py` fails on every Windows machine with `module 'socket' has no attribute 'AF_UNIX'`
- [Bug] Missing fallback when disable-model-invocation flag prevents mid-sentence dispatch recovery
- [BUG] "Clean my computer" skill deletes staged files on unrelated command
- Slash-command list does not pick up newly added .claude/commands skills until full app restart (Windows desktop)
- skill-creator: read_text()/write_text() sem encoding=utf-8 corrompem conteúdo acentuado no Windows
- Agent tool: skill path mentioned in subagent prompt auto-invokes the skill and can end the run with no work done
- [BUG] VS Code extension: namespaced /plugin:skill form shows “No matching commands” and cannot be submitted; skills whose name matches a built-in command are unreachable
- [BUG]
- Skill execution fails when glob pattern contains backticks - shell command substitution bug
- Custom skill not appearing in slash commands or Settings > Skills UI
- [BUG] [Tool Routing] Deferred MCP tools lack descriptions in ToolSearch, causing Skills with overlapping capabilities to always shadow them
- [BUG] Cowork: bundled Google Drive skill leads with a curl/script route that returns 401 (no credential in the sandbox shell); connector tools work
- [BUG] Mods: $.command.register refused for a user/project skill's name is undocumented, says "user's" for project skills, and the debug log omits the error
- / slash-command autocomplete: Down arrow past a large same-prefix block (gortex-*) causes unbounded re-loading
- [BUG] Desktop (Windows + WSL): new session screen doesn't load WSL skills or show Files/Git panels until the session starts
- [BUG] Cowork save_skill returns bare "Turn ended" with no success/error and installs nothing
- Skill(name) allow-list + bypassPermissions do not suppress the interactive skill-use prompt
- Skill tool resolves a stale plugin-cache copy of a user skill, silently, with no version signal
- Skill tool: one specific skill fails 'Unknown skill' while sibling skills installed in the same batch work fine
- [BUG] PreToolUse hook "if" field (permission-rule syntax) never matches Skill tool invocations (v2.1.278)
- [Bug] Session skill discovery doesn't reflect newly-added skills in already-enabled plugins
- [BUG]
- [BUG] Idle TUI CPU scales with the NUMBER of installed skills (~45 syscalls/s per skill; 40 skills ≈ 4–5 % of a core per idle session, Windows 2.1.276)
- `once: true` on skill-frontmatter hooks is ignored — the hook keeps firing after a successful run (2.1.273)
- File window silently fails to open a linked file outside the working directory
- /reload-skills reports "no changes" for a skill added to disk mid-session (Desktop app); a fresh process sees it
- [BUG] Desktop Code tab: personal skills missing from "/" autocomplete, and "/" typed mid-text shows no list (regression)