Control layer for LLM integrations that evaluates model output risks (SQL, command execution, etc.) before execution.
-
Updated
Apr 20, 2026 - Java
#
prompt-safety
Here are 7 public repositories matching this topic...
Field research exposing how LLM safeguards collapse under polite, persistent interaction. Includes full report, metrics, session logs, and the AION conditioning protocol.
ai-safety aion cognitive-systems security-research ai-ethics llm machine-conditioning prompt-safety resilience-failure ethics-fatigue
-
Updated
Sep 1, 2026 - Python
Field research exposing how LLM safeguards collapse under polite, persistent interaction. Includes full report, metrics, session logs, and the AION conditioning protocol.
ai-safety aion cognitive-systems security-research ai-ethics llm machine-conditioning prompt-safety resilience-failure ethics-fatigue
-
Updated
May 8, 2025 - Python
Discover, score, install, and audit Agent Skills for Codex & Claude Code with prompt-safety and supply-chain governance.
skills skill-management skill-discovery agent-skills prompt-safety claude-skills codex-skills skill-installer
-
Updated
Jul 3, 2026 - Python
Professional cross-agent privacy redaction and safe-sharing skill for text, screenshots, resumes, logs, prompts, forms, support messages, and public posts.
-
Updated
May 16, 2026 - PowerShell
Intent alignment guard for Claude Code — 5-question task-start protocol, tiered action gate, and PreToolUse hook that blocks destructive actions before they execute.
ai-safety claude ai-agent anthropic llm-security agentic-ai claude-code prompt-safety intent-alignment claude-skills
-
Updated
Jul 26, 2026 - Python
Local pre-send risk guard for Claude Code prompts with safe rewrites and audit reports.
-
Updated
Apr 11, 2026 - Python
Add this topic to your repo
To associate your repository with the prompt-safety topic, visit your repo's landing page and select "manage topics."