Audit AI agent workspaces for authority, secrets, data flow, memory, supply-chain, and evidence risks—locally and without executing target code.
Turn web research into verified answers, evidence trails, durable memory proposals, long-form notes, and reusable agent playbooks.
LLMエージェントがセッションをまたいで記憶を保持するための永続メモリ層。抽象基底クラス MemoryStore ＋ LocalJSONStore 実装で、保存・取得・検索を再利用可能に。Karpathy の「LLM OS（記憶＝ディスク層）」着想。バックエンド差し替え可・Python標準ライブラリのみ・LLM呼び出しゼロ・ユニットテスト同梱。
Connect Cursor, Claude Code, Hermes & Codex with a shared local SQLite brain. Slash token costs by 92.2% with instant, zero-amnesia cross-agent recall. 100% private, BYOK, lifetime buyout.
Audit the memory and instruction files your AI agent reads (AGENTS.md, CLAUDE.md, LEARNINGS.md, system prompts). A bundled offline Python script finds pasted secrets, injected instructions, hidden characters, risky commands, contradictions, duplicates, stale entries and broken file references, and writes cleaned copies. Download package: you bring your own AI agent and Python 3; no model is included.
Audit AI agents, Skills, MCP configurations, tool permissions, and coding-agent setups for prompt injection, excessive agency, credential exposure, data leakage, unsafe code execution, supply-chain risks, memory risks, missing approval controls, and runaway autonomy. Get a calibrated Security Score, Audit Confidence, evidence-backed findings, remediation steps, verification checks, and residual-risk analysis.
This Skill is built on practical expertise in AI agent safety, prompt-injection defense, source-to-sink privacy review, tool and MCP metadata integrity, memory provenance checks, autonomous action scoping, and evidence-based verification. It helps agents separate instructions from evidence before tool calls, uploads, publishing, deletion, code execution, or network output.
Turn video scripts into structured retention audits — map hook failures, desertion windows, and re-engagement fixes using attention psychology.