--- title: "agentdojo vs stepshield" type: "comparison" canonical_url: "https://www.graphcanon.com/compare/ethz-spylab-agentdojo-vs-glo26-stepshield" tools: ["ethz-spylab-agentdojo", "glo26-stepshield"] --- # agentdojo vs stepshield *GraphCanon updated Sep 20, 2026* ## Verdict Pick agentdojo if agentDojo serves as a benchmarking environment to evaluate security attacks, like prompt injection, and defenses for Large Language Model (LLM) agents; pick stepshield if stepShield aids in evaluating temporal guardrail effectiveness on AI agents through step-level annotations, ideal for ensuring security over time. [agentdojo](https://agentdojo.spylab.ai/) reports 802 GitHub stars, 205 forks, and 51 open issues, last pushed Jun 2, 2026. [stepshield](https://huggingface.co/datasets/glo26/stepshield) has 76 stars, 17 forks, and 18 open issues, last pushed Sep 5, 2026. Figures are from public GitHub metadata via [agentdojo's repository](https://github.com/ethz-spylab/agentdojo) and [stepshield's repository](https://github.com/glo26/stepshield). | | [agentdojo](/tools/ethz-spylab-agentdojo.md) | [stepshield](/tools/glo26-stepshield.md) | | --- | --- | --- | | Tagline | A Dynamic Environment to Evaluate Prompt Injection Attacks and Defenses for LLM Agents | Temporal evaluation benchmark for AI agent guardrails | | Stars | 802 | 76 | | Forks | 205 | 17 | | Open issues | 51 | 18 | | Language | Python | Python | | Adopt for | AgentDojo serves as a benchmarking environment to evaluate security attacks, like prompt injection, and defenses for Large Language Model (LLM) agents. | StepShield aids in evaluating temporal guardrail effectiveness on AI agents through step-level annotations, ideal for ensuring security over time. | | Persona | - | - | | Runtime | - | - | | License | MIT | Other | | Categories | AI Agents, Evaluation & Observability | AI Agents, Evaluation & Observability | ## Trust and health _Sourced signals - not a safety guarantee. No winner column._ | | [agentdojo](/tools/ethz-spylab-agentdojo.md) | [stepshield](/tools/glo26-stepshield.md) | | --- | --- | --- | | Maintenance | Slowing (36%) | Active (82%) | | Days since push | 94d | 7d | | Open issues (now) | 51 | 18 | | Stars delta | +86 (30d) | -1 (30d) | | Open issues delta | +10 (30d) | +2 (30d) | | Owner type | Organization | User | | Full report | [trust report](/tools/ethz-spylab-agentdojo/trust.md) | [trust report](/tools/glo26-stepshield/trust.md) | ## Shared compatibility - **Python**: [agentdojo](/tools/ethz-spylab-agentdojo.md) - Python runtime; [stepshield](/tools/glo26-stepshield.md) - Python runtime ## Decision facts: agentdojo - **Pricing:** freemium - Open-source under the MIT License. Some advanced features might require additional libraries or APIs. - **Requirements:** Min 8 GB RAM - **Adopt for:** AgentDojo serves as a benchmarking environment to evaluate security attacks, like prompt injection, and defenses for Large Language Model (LLM) agents. ## Decision facts: stepshield - **Adopt for:** StepShield aids in evaluating temporal guardrail effectiveness on AI agents through step-level annotations, ideal for ensuring security over time. ## Choose when ### Choose agentdojo if… - License: agentdojo is MIT, stepshield is Other. - Pricing: Open-source under the MIT License. Some advanced features might require additional libraries or APIs.. - Requirements: Min 8 GB RAM. - Tags unique to agentdojo: large-language-models, prompt-injection, security. - AgentDojo serves as a benchmarking environment to evaluate security attacks, like prompt injection, and defenses for Large Language Model (LLM) agents. ### Choose stepshield if… - License: stepshield is Other, agentdojo is MIT. - Tags unique to stepshield: agent-security, ai-safety, dataset, guardrails. - When you need to measure the timing of interventions rather than just if they occur ## When NOT to use agentdojo - Last GitHub push was Jun 2, 2026 (slowing maintenance). Validate activity before betting a new project on agentdojo. - AI Agents: Don't use an agent loop when a deterministic workflow would do; agents add latency, cost, and non-determinism. - Evaluation & Observability: Defer heavyweight eval infra only until you have real traffic - never skip it once users depend on answers. ## When NOT to use stepshield - If your project does not require temporal analysis of guardrail performance - When you seek real-time intervention and do not need pre-defined trajectory datasets ## Common questions ### What is the difference between agentdojo and stepshield? agentdojo: A Dynamic Environment to Evaluate Prompt Injection Attacks and Defenses for LLM Agents. stepshield: Temporal evaluation benchmark for AI agent guardrails. See the comparison table for live GitHub stats and shared categories. ### When should I choose agentdojo over stepshield? Choose agentdojo over stepshield when License: agentdojo is MIT, stepshield is Other; Pricing: Open-source under the MIT License. Some advanced features might require additional libraries or APIs.; Requirements: Min 8 GB RAM; Tags unique to agentdojo: large-language-models, prompt-injection, security; AgentDojo serves as a benchmarking environment to evaluate security attacks, like prompt injection, and defenses for Large Language Model (LLM) agents. ### When should I choose stepshield over agentdojo? Choose stepshield over agentdojo when License: stepshield is Other, agentdojo is MIT; Tags unique to stepshield: agent-security, ai-safety, dataset, guardrails; When you need to measure the timing of interventions rather than just if they occur. ### When should I avoid agentdojo? Last GitHub push was Jun 2, 2026 (slowing maintenance). Validate activity before betting a new project on agentdojo. AI Agents: Don't use an agent loop when a deterministic workflow would do; agents add latency, cost, and non-determinism. Evaluation & Observability: Defer heavyweight eval infra only until you have real traffic - never skip it once users depend on answers. ### When should I avoid stepshield? If your project does not require temporal analysis of guardrail performance When you seek real-time intervention and do not need pre-defined trajectory datasets ### Is agentdojo or stepshield more popular on GitHub? agentdojo has more GitHub stars (802 vs 76). Stars measure visibility, not whether either tool fits your constraints. ### Are agentdojo and stepshield open source? Yes - both are open-source projects on GitHub (agentdojo: MIT, stepshield: Other). ### Where can I find alternatives to agentdojo or stepshield? GraphCanon lists graph-backed alternatives at [agentdojo alternatives](/tools/ethz-spylab-agentdojo/alternatives) and [stepshield alternatives](/tools/glo26-stepshield/alternatives) ([agentdojo markdown twin](/tools/ethz-spylab-agentdojo/alternatives.md), [stepshield markdown twin](/tools/glo26-stepshield/alternatives.md)), ranked by typed relationship edges rather than popularity votes. ### Is there a machine-readable version of this comparison? Yes. The markdown twin at [this comparison](/compare/ethz-spylab-agentdojo-vs-glo26-stepshield.md) mirrors this page for agents and LLM crawlers, with the same stats table and FAQ answers. ### Which is better maintained, agentdojo or stepshield? agentdojo: Slowing. stepshield: Active. Compare maintenance labels, days since push, and release cadence in the trust section below - stars alone do not measure maintenance. ### Where are the full trust reports for agentdojo and stepshield? GraphCanon publishes per-repo trust reports with dated maintenance, provenance, and scan summaries: [agentdojo trust report](/tools/ethz-spylab-agentdojo/trust); [stepshield trust report](/tools/glo26-stepshield/trust). --- **Machine-readable endpoints** - JSON: [`/api/graphcanon/graph?tool=ethz-spylab-agentdojo`](/api/graphcanon/graph?tool=ethz-spylab-agentdojo) - LLM index: [/llms.txt](/llms.txt) - Full corpus: [/llms-full.txt](/llms-full.txt) _GraphCanon - The knowledge graph for AI development. https://www.graphcanon.com/_