By ddaanet
Forces `git status` to run with the Claude Code command sandbox disabled, so it reports the real working tree instead of phantom sandbox-mount entries.
A Claude Code plugin that forces git status to run with the command
sandbox disabled.
Under Claude Code's command sandbox, mounted paths (notably the agent's
.claude directory) surface as phantom entries in a repository's working
tree — character-device nodes that exist only inside the sandbox view. A
git status run inside the sandbox sees them and reports files that are
not actually there, so the agent reads a polluted, untrustworthy status.
The same command run unsandboxed reports the real working tree.
This plugin closes that gap mechanically: it intercepts git status
before it runs and, if the call is sandboxed, denies it with an
instruction to re-run the identical command with the sandbox off.
PreToolUse(Bash) hook inspects every Bash command.git status invocation and the call is sandboxed,
it denies the call and tells the agent to re-run with
dangerouslyDisableSandbox: true.git status that is already unsandboxed.It is precision-biased: it would rather let an exotic invocation through
than block an unrelated command. git commit -m "fix status bar",
git log --grep status, and echo "git status" all pass.
/plugin marketplace add ddaanet/claude-plugins
/plugin install unsandbox-git-status@ddaanet
/reload-plugins
Replace the marketplace ref with wherever this plugin is published. After
installing, the hook is active in new sessions (run /reload-plugins to
pick it up in the current one).
When it fires you will see the call denied with a reason like:
git status was run sandboxed. Re-run the exact same Bash command with dangerouslyDisableSandbox set to true.
The agent then re-issues the command unsandboxed and gets the true status.
bash, jq.
just precommit # shellcheck + syntax + the hook test
just release minor
See DESIGN.md for the rationale and CLAUDE.md for agent-facing
working notes. Release tooling is the vendored
claude-plugin-dev
toolkit under plugin-dev/.
MIT
Executes bash commands
Hook triggers when Bash tool is used
Own this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimOwn this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimBased on adoption, maintenance, documentation, and repository signals. Not a security audit or endorsement.
npx claudepluginhub ddaanet/claude-plugins --plugin unsandbox-git-statusPre-/clear task snapshot for Claude Code: agent writes current task + open decisions; at next session start, a hook scrapes the prior transcript and injects task + last-N prompts + files touched into the fresh agent automatically.
Opinionated agent framework for Claude Code
Candidature assistée : préparation, lettre de motivation, CV adapté, relecture, suivi des retours. Stockage Notion. Contenu français.
Shell scripting gotchas skill plus automatic shellcheck feedback on edited shell files
Commit-msg git hook that replaces conventional-commit prefixes (feat:, fix:, docs:, ...) with gitmoji emojis. Installs the per-repo hook automatically on SessionStart whenever the plugin is enabled in the current repo.
MCP server that saves 98% of your context window with session continuity. Sandboxed code execution in 11 languages, FTS5 knowledge base with BM25 ranking, and automatic state restore across compactions.
Complete AI coding workflow system. Self-correcting memory + persistent FTS5-indexed research wikis + auto-research loop + multi-LLM council on a single SQLite store. 33 skills, 8 agents, 22 commands, 37 hook scripts across 24 events. Cross-agent via SkillKit.
Open-source, local-first Claude Code plugin for token reduction, context compression, and cost optimization using hybrid RAG retrieval (BM25 + vector search), reranking, AST-aware chunking, and compact context packets.
Comprehensive skill pack with 66 specialized skills for full-stack developers: 12 language experts (Python, TypeScript, Go, Rust, C++, Swift, Kotlin, C#, PHP, Java, SQL, JavaScript), 10 backend frameworks, 6 frontend/mobile, plus infrastructure, DevOps, security, and testing. Features progressive disclosure architecture for 50% faster loading.
Persistent file-based planning for AI coding agents. Crash-proof markdown plans (task_plan.md, findings.md, progress.md) that survive context loss and /clear, with an opt-in completion gate and multi-agent shared state. Manus-style. Works with Claude Code, Codex CLI, Cursor, Kiro, OpenCode and 60+ agents via the SKILL.md standard. Includes Arabic, German, Spanish, and Chinese (Simplified and Traditional).