{"data":{"slug":"glo26-stepshield","name":"stepshield","tagline":"Temporal evaluation benchmark for AI agent guardrails","github_url":"https://github.com/glo26/stepshield","owner":"glo26","repo":"stepshield","owner_avatar_url":"https://avatars.githubusercontent.com/u/15223526?v=4","primary_language":"Python","stars":76,"forks":17,"topics":["agent-security","ai-safety","benchmark","dataset","guardrails","neurips-2026","step-level-detection"],"archived":false,"github_pushed_at":"2026-09-05T20:01:00+00:00","maintenance_label":"Active","stars_delta_30d":-1,"url":"https://www.graphcanon.com/tools/glo26-stepshield","markdown_url":"https://www.graphcanon.com/tools/glo26-stepshield.md","api_url":"https://www.graphcanon.com/api/graphcanon/tools/glo26-stepshield","graph_url":"https://www.graphcanon.com/api/graphcanon/graph?tool=glo26-stepshield","description":"StepShield: When, Not Whether to Intervene on Rogue Agents — NeurIPS 2026 benchmark for temporal evaluation of AI agent guardrails (9,429 trajectories, step-level annotations)","homepage_url":"https://huggingface.co/datasets/glo26/stepshield","license":"Other","open_issues":18,"watchers":8,"ai_summary":"StepShield provides trajectories and step-level annotations to evaluate the temporal intervention effectiveness required for securing rogue AI agents.","readme_excerpt":"### 1. Install\n\n```bash\ngit clone https://github.com/glo26/stepshield.git\ncd stepshield\npip install -r requirements.txt\n```\n\n---\n\n## License\n\nThe code in this repository is licensed under the [MIT License](LICENSE). The dataset is licensed under [CC BY 4.0](https://creativecommons.org/licenses/by/4.0/).\n\nThis project was supported by the **MOVE Fellowship**. For questions or collaboration inquiries, contact `contact@stepshield.ai`.","github_created_at":"2025-12-08T05:26:03+00:00","created_at":"2026-07-15T10:44:01.14159+00:00","updated_at":"2026-09-20T04:26:15.372082+00:00","categories":[{"slug":"ai-agents","name":"AI Agents","url":"https://www.graphcanon.com/categories/ai-agents","markdown_url":"https://www.graphcanon.com/categories/ai-agents.md","api_url":"https://www.graphcanon.com/api/graphcanon/categories/ai-agents"},{"slug":"evaluation-observability","name":"Evaluation & Observability","url":"https://www.graphcanon.com/categories/evaluation-observability","markdown_url":"https://www.graphcanon.com/categories/evaluation-observability.md","api_url":"https://www.graphcanon.com/api/graphcanon/categories/evaluation-observability"}],"tags":[{"slug":"agent-security","name":"agent-security"},{"slug":"ai-safety","name":"ai-safety"},{"slug":"benchmark","name":"benchmark"},{"slug":"dataset","name":"dataset"},{"slug":"guardrails","name":"guardrails"},{"slug":"neurips-2026","name":"neurips-2026"},{"slug":"step-level-detection","name":"step-level-detection"}],"trust":{"provenance":{"is_fork":false,"github_id":1112068019,"owner_type":"User","methodology":"github_public_v1","parent_repo":null,"near_duplicate_slugs":[]},"computed_at":"2026-09-13T06:00:05.750Z","maintenance":{"label":"Active","score":82,"methodology":"github_public_v1","releases_90d":0,"days_since_push":7,"last_release_at":null,"stars_delta_30d":-1,"open_issues_delta_30d":2},"security_summary":{"status":"findings","scanner":"osv@v1","low_count":6,"high_count":0,"last_scan_at":"2026-07-15T10:44:02.585Z","medium_count":0,"scan_profile":"deps","critical_count":0}},"capability_facts":{"scan":{"source":"repo_scan","observed_at":"2026-09-13T06:00:06.463Z"},"languages":{"value":["python"],"source":"github.language+pyproject.toml","observed_at":"2026-09-13T06:00:06.463Z"},"license_spdx":{"value":"Other","source":"github.license","observed_at":"2026-09-13T06:00:06.463Z"}},"decision_facts":{"hosting":null,"pricing":null,"requirements":null,"constraints":null,"when_to_use":["When you need to measure the timing of interventions rather than just if they occur","For benchmarking your AI safeguards against a curated dataset of 9,429 trajectories"],"when_not_to_use":["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"],"source":"enrich:decision_facts","observed_at":"2026-07-16T18:42:12.919Z"},"constraint_facets":null,"decision_summary":[{"label":"Adopt for","value":"StepShield aids in evaluating temporal guardrail effectiveness on AI agents through step-level annotations, ideal for ensuring security over time."}]}}