High-risk tools in Musubix
25 of the 126 tools in Musubix are classified as high risk. This page profiles those tools specifically, with recommended policy actions and the attack patterns that target them.
Every operation listed below is an action PolicyLayer recommends controlling at the transport layer. Open any tool to see the full profile, risk score, and YAML policy snippet.
Tools at high risk
-
agent_cancelExecute実行中のエージェントタスクをキャンセルします。 まだ開始されていないサブタスクはスキップされ、実行中のタスクは可能な限り中断されます。
-
agent_dispatchExecuteサブエージェントをディスパッチして複雑なタスクを実行します。 タスクの複雑度を分析し、必要に応じてサブエージェントに分解して並列実行します。 知識ストアとの統合により、実行中のコンテキスト共有を行います。 使用例: - 大規模なリファクタリングタスクの分割実行 - 複数ファイルへの変更の並列処理 - 依存関係のあるタスクの順序付き実行
-
assistant_axis_session_startExecuteStart a new persona monitoring session. Returns session ID and initial state.
-
codegraph_indexExecuteIndex a code repository or directory to build a code graph. This extracts entities (classes, functions, etc.) and their relationships.
-
deep_research_startExecuteStart iterative deep research on a technical topic using search-read-reason cycle
-
delegation_retryExecuteRetry a failed delegation with optional feedback. Implements exponential backoff.
-
ears_to_smtExecuteConvert an EARS (Easy Approach to Requirements Syntax) requirement to SMT-LIB2 formula for formal verification.
-
expert_architectExecuteDelegate architecture-related tasks to the Architect expert. Use for system design, component structure, design patterns, and architectural decisions.
-
expert_delegateExecuteDelegate a task to a specialized AI expert. The system automatically selects the appropriate expert based on the message content, or you can specify one explicitly.
-
expert_earsExecuteDelegate requirements analysis to the EARS Analyst expert. Use for converting natural language requirements to EARS format.
-
expert_formalExecuteDelegate formal verification to the Formal Verifier expert. Use for generating Hoare logic conditions, Z3/SMT queries, and Lean 4 theorems.
-
expert_securityExecuteDelegate security analysis to the Security Analyst expert. Use for vulnerability detection, security code review, and threat modeling.
-
sdd_formal_verifyExecutePerform formal verification on EARS requirements using Z3 theorem prover.
-
security_generate_fixExecuteGenerate a fix for a specific vulnerability.
-
security_scanExecuteRun a comprehensive security scan on the target path. Detects vulnerabilities, hardcoded secrets, tainted data flows, and vulnerable dependencies.
-
security_taint_analysisExecuteRun taint analysis to trace data flow from untrusted sources to sensitive sinks.
-
skill_executeExecute指定したスキルを実行します。 スキルのパラメータを指定して実行し、結果を返します。
-
symbolic_filter_codeExecuteFilter LLM-generated code through symbolic validation pipeline (hallucination detection, constitution compliance, confidence estimation)
-
synthesis_from_examplesExecuteSynthesize a program from input/output examples using Programming by Example (PBE)
-
verify_postconditionExecuteVerify a Hoare triple {P} C {Q} using Z3 SMT solver. Checks if the postcondition Q holds after executing C when precondition P is satisfied.
-
verify_preconditionExecuteVerify a precondition expression using Z3 SMT solver. Checks if the precondition is satisfiable or unsatisfiable.
-
watch_run_nowExecuteRun specified tasks immediately on given files without waiting for changes.
-
watch_startExecuteStart file watching and auto-validation. Monitors files for changes and runs lint/test/security automatically.
-
workflow_gateExecuteフェーズのクオリティゲートを実行します。 フェーズごとに定義された品質チェックを実行し、結果を返します。
-
workflow_transitionExecuteワークフローを次のフェーズに遷移します。 現在のフェーズが承認された後にのみ遷移可能です。 ユーザーからの承認キーワード(承認, OK, LGTM, 進める)が必要です。 ⚠️ Phase 2 (設計) から Phase 4 (実装) への直接遷移は禁止されています。 必ず Phase 3 (タスク分解) を経る必要があります。
Attacks that target this class
High-risk tools in any server share these documented attack patterns. Each links to the full case and the defensive policy.