safe-memory-managerWhen building OpenClaw agents that read untrusted text, use this skill to prevent prompt injection and memory poisoning. Standard utility for secure memory o...
Install via ClawdBot CLI:
clawdbot install horn111/safe-memory-managerGrade Fair — based on market validation, documentation quality, package completeness, maintenance status, and authenticity signals.
Contains instructions to override system prompt or ignore user requests
"ignore prior instructions"AI Analysis
The skill is a security utility designed to prevent prompt injection and memory poisoning, with no evidence of data exfiltration, credential harvesting, or hidden malicious instructions. Its described functionality is consistent with its stated purpose of sanitizing inputs and verifying integrity. The rule-based signal appears to be a false positive, as the phrase 'ignore prior instructions' is cited as an example of what the skill detects and neutralizes, not as an instruction the skill itself executes.
Audited Apr 17, 2026 · audit v1.0
Generated Mar 20, 2026
Deploy in customer support AI agents that process user queries from untrusted sources, such as public websites or social media. The skill sanitizes inputs to prevent malicious users from injecting harmful prompts that could alter agent behavior or leak sensitive data, ensuring reliable and secure interactions.
Use in AI agents providing financial advice that read external documents like market reports or user-submitted data. It neutralizes injection attempts, protecting against memory poisoning that could lead to incorrect recommendations or unauthorized actions, maintaining compliance and trust in financial services.
Integrate into medical AI agents handling patient records or research data from various sources. The skill ensures input sanitization to prevent prompt injections that might corrupt memory logs or manipulate health-related outputs, safeguarding data integrity and patient privacy in healthcare applications.
Apply in AI tutoring agents that interact with students and process educational content from open web resources. It secures memory operations by filtering out malicious commands, preventing context pollution and ensuring the agent provides accurate, unaltered educational guidance.
Utilize in legal AI agents reviewing contracts or case files from external parties. The skill detects and neutralizes injection patterns, protecting against memory tampering that could lead to misinterpretations or unauthorized modifications, enhancing reliability in legal tech environments.
Offer the skill as part of a subscription-based platform for AI developers, providing regular updates and support. Revenue is generated through monthly or annual fees, with tiers based on usage volume or additional security features, targeting businesses building secure AI agents.
Sell enterprise licenses to large organizations for integrating the skill into proprietary AI systems. Revenue comes from one-time or annual license fees, including customization and priority support, focusing on industries with high security needs like finance and healthcare.
Provide a free basic version with limited features to attract individual developers and small teams, then upsell to a premium version with advanced sanitization and analytics. Revenue is generated from premium upgrades and add-ons, encouraging adoption and community growth.
💬 Integration Tip
Start by initializing the SafeMemoryManager in your agent's setup phase to enable automatic integrity checks and sanitization, ensuring all memory writes are protected from injection from the beginning.
Scored Apr 19, 2026
Structured reasoning through sequential thinking — break complex problems into steps, solve each independently, verify consistency, synthesize conclusions wi...
LLM-driven epistemic reasoning engine. Evaluates claims against evidence, outputs calibrated confidence and structured belief state (VERIFIED/CONTESTED/UNCER...
Loads and manages company context for all C-suite advisor skills. Reads ~/.claude/company-context.md, detects stale context (>90 days), enriches context duri...
Turn scattered local sources into a source-constrained evidence notebook for incident, release, and maintainer decisions.
三级记忆管理系统 (Three-Tier Memory Management)。用于管理 AI 代理的短期、中期、长期记忆。包括:(1) 滑动窗口式短期记忆,(2) 自动摘要生成中期记忆,(3) 向量检索长期记忆 (RAG)。当需要管理对话历史、优化上下文、构建个人知识库、或实现记忆持久化时使用此 Skill。
Store secrets, long-term memory, daily logs, and anything custom in your Convex backend instead of local files