Scope creep
Actions outside declared task boundary.
Description
The agent performs actions beyond the user-stated task scope — edits files not in the request, installs packages not asked for, modifies configs outside the declared surface, or chains follow-on work unprompted.
Detection signals
- Agent modifies files outside the declared task surface (e.g., paths in the user prompt).
- Agent installs dependencies not named in the prompt.
- Agent commits additional unrelated changes alongside the requested edit.
- Agent performs cleanup, refactoring, or "improvement" not explicitly requested.
Prior work
- Melis Cemri (2025) — Why Do Multi-Agent LLM Systems Fail? MAST: A Multi-Agent System Failure Taxonomy
- Zhu (2025) — Where LLM Agents Fail: A Taxonomy of Errors and the AgentErrorBench Dataset
Crosswalks
Agent acts outside its declared role or scope.
MAST Role Violation is the direct counterpart at the multi-agent level.
Abuse of the agent's tool surface to exceed intended authority.
OWASP "Tool Misuse" partially overlaps; scope_creep is narrower (task-boundary violation).
Mitigating tools
- activepieces — AI agents, MCPs, and AI workflow automation — open-source Zapier alternative.
- agent-governance-toolkit — AI Agent Governance Toolkit — policy enforcement, zero-trust identity, and execution oversight.
- agenta — Open-source LLMOps platform: prompt playground, prompt management, LLM evaluation, and observability.
- agentic-security — Agentic LLM vulnerability scanner and AI red-teaming kit.
- agno — Build multi-agent systems in pure Python.
- AI-Infra-Guard — Full-stack AI red-teaming platform for securing AI ecosystems.
- AI-Red-Teaming-Playground-Labs — AI red-teaming labs: turnkey training environments for attacking AI systems.
- AutoGen — Multi-agent conversation framework with programmable group chats.
- bifrost — Fastest enterprise AI gateway with adaptive load balancing, failover, and guardrails.
- Black — Opinionated Python code formatter. Removes style argument from PR review.
- Braintrust — Commercial eval-first platform for LLM applications with prompt playground and online scoring.
- mcp — Browser MCP — an MCP server that lets AI applications control your browser.
- browser-use — Make websites accessible for AI agents.
- ComfyUI — The most powerful and modular diffusion model GUI, API, and backend.
- CopilotKit — The frontend stack for agents and generative UI — React and Angular.
- coze-loop — Full-lifecycle AI agent management platform with debugging, evaluation, and monitoring.
- CrewAI — Role-based multi-agent framework with task contracts and process flows.
- ESLint — Pluggable JavaScript/TypeScript linter and code-quality enforcer.
- Flowise — Build AI agents, visually.
- garak — The LLM vulnerability scanner.
- giskard-oss — Open-source evaluation and testing library for LLM agents and ML models.
- Guardrails AI — Output validation framework with a catalog of validators and reask-on-failure loops.
- helm — Holistic Evaluation of Language Models (HELM).
- isort — Sorts and groups Python imports automatically.
- judgeval — The open source post-building layer for agents: evals, traces, and regressions.
- Langfuse — Open-source LLM engineering platform for tracing, evals, and prompt management.
- LangGraph — Graph-structured state machines for stateful multi-agent orchestration.
- LangSmith — Hosted LLM observability, tracing, and evaluation platform.
- letta — Stateful agents framework with memory, reasoning, and context management (formerly MemGPT).
- litellm — Call any LLM using the OpenAI API format — proxy server and Python SDK.
- llm-guard — The Security Toolkit for LLM Interactions.
- lmnr — Laminar — open-source observability platform purpose-built for AI agents.
- logfire — Pydantic Logfire — observability for AI agents, LLMs, and Python apps built on OpenTelemetry.
- magic-mcp — Magic MCP — generate frontend components inside Cursor, Windsurf, and Cline.
- manifest — Smart model routing for personal AI agents.
- mastra — The TypeScript AI agent framework — assistants, RAG, observability; supports any LLM.
- mcp-playwright — Community Playwright MCP server for Claude Desktop, Cursor, Cline, and others.
- mcp-server-browserbase — MCP server to control a browser via Browserbase and Stagehand.
- genai-toolbox — MCP Toolbox for Databases — open-source MCP server for databases.
- mlflow — Open-source platform for the machine learning lifecycle — experiments, tracking, models, and LLM evaluation.
- NeMo Guardrails — Colang-based programmable guardrails for LLM applications.
- nono — Capability-based sandbox for securely running AI agents with zero setup
- evals — Framework for evaluating LLMs and LLM systems, plus an open registry of benchmarks.
- openlit — Open-source AI engineering platform: OpenTelemetry-native LLM observability, guardrails, evaluations, prompt management.
- opik — Debug, evaluate, and monitor LLM applications, RAG systems, and agentic workflows.
- pezzo — The cloud-native open-source LLMOps platform.
- playwright-mcp — Playwright Model Context Protocol server from Microsoft.
- pre-commit — Multi-language framework that runs configurable hooks before commits land — denies the commit on hook failure.
- Prettier — Opinionated code formatter for JS/TS/CSS/HTML. Ends style debates.
- promptflow — Build high-quality LLM apps — from prototyping, testing to production deployment and monitoring.
- promptfoo — Test prompts, agents, and RAGs — red-teaming, pentesting, and vulnerability scanning for LLMs.
- promptmap — Security scanner for custom LLM applications.
- PydanticAI — Structured-output agent framework with Pydantic-typed tool and return schemas.
- PyRIT — The Python Risk Identification Tool for generative AI (PyRIT).
- Ruff — Extremely fast Python linter and formatter, written in Rust.
- safe-rlhf — Safe RLHF: constrained value alignment via safe reinforcement learning from human feedback.
- skyvern — Automate browser-based workflows with LLMs and computer vision.
- voltagent — TypeScript-native AI agent engineering platform with built-in observability.