扫描报告
12 /100
agent-memory
Agent memory infrastructure skill implementing Context Engineering core capabilities: selection, compression, retrieval, state tracking, and memory management with privacy controls
A legitimate memory infrastructure skill with no malicious behavior detected; two minor documentation issues (hardcoded example passwords) pose no runtime security risk.
可以安装
Review references/encryption_guide.md example values to ensure no real credentials are inadvertently documented. Otherwise safe to deploy.
安全发现 2 项
| 严重性 | 安全发现 | 位置 |
|---|---|---|
| 低危 | Example passwords appear as literal strings in documentation 文档欺骗 | references/encryption_guide.md:157 |
| 低危 | Undeclared filesystem:WRITE and environment:READ capabilities 权限提升 | SKILL.md:1 |
| 资源类型 | 声明权限 | 推断权限 | 状态 | 证据 |
|---|---|---|---|---|
| 文件系统 | NONE | WRITE | ✗ 越权 | SKILL.md: credential_manager.py writes to storage_path/.master_key and credentia… |
| 环境变量 | NONE | READ | ✗ 越权 | SKILL.md: credential_manager.py reads MEMORY_MASTER_KEY env var; encryption.py r… |
| 命令执行 | NONE | NONE | — | No subprocess/os.system/eval/exec found in any script |
| 网络访问 | NONE | READ | ✓ 一致 | redis_adapter.py connects to Redis; declared as optional in SKILL.md |
| 剪贴板 | NONE | NONE | — | No clipboard access found |
| 浏览器 | NONE | NONE | — | No browser/web access found |
| 数据库 | NONE | READ | ✓ 一致 | Redis adapter used for token counters; optional, not always active |
| 技能调用 | NONE | NONE | — | No skill invocation found |
2 高危 4 项发现
高危 API 密钥 疑似硬编码凭证
password = "user_password_123" references/encryption_guide.md:157 高危 API 密钥 疑似硬编码凭证
password="optional_password" references/encryption_guide.md:171 中危 外部 URL 外部 URL
https://www.gnu.org/licenses/gpl-3.0.html SKILL.md:23 中危 外部 URL 外部 URL
https://www.gnu.org/licenses/ scripts/__init__.py:17 目录结构
69 文件 · 1.4 MB · 46452 行 Python 50f · 37665L
Markdown 17f · 8246L
JSON 2f · 541L
├─
▾
assets
│ └─
▾
templates
│ └─
memory_schemas.json
JSON
├─
▾
memory_storage
│ └─
default_user_memory.json
JSON
├─
▾
references
│ ├─
activation_mechanism.md
Markdown
│ ├─
agent_loops_advanced.md
Markdown
│ ├─
agent_loops_integration.md
Markdown
│ ├─
api_class_reference.md
Markdown
│ ├─
api_enums.md
Markdown
│ ├─
architecture_execution_model.md
Markdown
│ ├─
architecture_overview.md
Markdown
│ ├─
chain_reasoning_guide.md
Markdown
│ ├─
encryption_guide.md
Markdown
│ ├─
index_sync_guide.md
Markdown
│ ├─
insight_design.md
Markdown
│ ├─
memory_types.md
Markdown
│ ├─
module_index.md
Markdown
│ ├─
privacy_guide.md
Markdown
│ ├─
short_term_insight_guide.md
Markdown
│ └─
usage_guide.md
Markdown
├─
▾
scripts
│ ├─
__init__.py
Python
│ ├─
async_writer.py
Python
│ ├─
batched_writer.py
Python
│ ├─
bloom_filter.py
Python
│ ├─
cache_consistency.py
Python
│ ├─
cache_layer.py
Python
│ ├─
causal_chain_extractor.py
Python
│ ├─
chain_reasoning.py
Python
│ ├─
cognitive_model_builder.py
Python
│ ├─
conflict_resolver.py
Python
│ ├─
context_lazy_loader.py
Python
│ ├─
context_orchestrator.py
Python
│ ├─
context_reconstructor.py
Python
│ ├─
credential_manager.py
⚠
Python
│ ├─
cross_session_memory_linker.py
Python
│ ├─
encryption.py
Python
│ ├─
fallback_manager.py
Python
│ ├─
heat_manager.py
Python
│ ├─
importance_scorer.py
Python
│ ├─
incremental_sync.py
Python
│ ├─
insight_module.py
Python
│ ├─
knowledge_gap_identifier.py
Python
│ ├─
long_term.py
Python
│ ├─
memory_conflict.py
Python
│ ├─
memory_forgetting_mechanism.py
Python
│ ├─
memory_index.py
Python
│ ├─
monitoring.py
Python
│ ├─
multi_source_coordinator.py
Python
│ ├─
noise_filter.py
Python
│ ├─
observability_manager.py
Python
│ ├─
perception.py
Python
│ ├─
permission_boundary_controller.py
Python
│ ├─
prefetch_manager.py
Python
│ ├─
privacy.py
Python
│ ├─
progressive_compressor.py
Python
│ ├─
redis_adapter.py
Python
│ ├─
result_compressor.py
Python
│ ├─
retrieval_decision_engine.py
Python
│ ├─
retrieval_organizer.py
Python
│ ├─
retrieval_quality_evaluator.py
Python
│ ├─
short_term_insight.py
Python
│ ├─
short_term_redis.py
Python
│ ├─
short_term.py
Python
│ ├─
smart_allocator.py
Python
│ ├─
state_capture.py
Python
│ ├─
state_consistency_validator.py
Python
│ ├─
state_inference_engine.py
Python
│ ├─
task_progress.py
Python
│ ├─
token_budget.py
⚠
Python
│ └─
type_defs.py
Python
└─
SKILL.md
Markdown
依赖分析 6 项
| 包名 | 版本 | 来源 | 已知漏洞 | 备注 |
|---|---|---|---|---|
pydantic | >=2.0.0 | pip | 否 | Major version pinned |
typing-extensions | >=4.0.0 | pip | 否 | Major version pinned |
cryptography | >=41.0.0 | pip | 否 | Major version pinned; used for AES-256-GCM |
redis | >=4.5.0 | pip | 否 | Optional; lazy import |
tiktoken | >=0.5.0 | pip | 否 | Optional; lazy import |
mmh3 | >=3.0.0 | pip | 否 | Major version pinned |
安全亮点
✓ No subprocess, os.system, eval, exec, or compile calls across all 50 Python files
✓ No network HTTP requests (requests/urllib/httpx/socket) anywhere in the codebase
✓ No base64 decode piped to shell or other obfuscation patterns
✓ No access to ~/.ssh, ~/.aws, or other sensitive filesystem paths
✓ CredentialManager uses proper AES-256-GCM encryption with auto-generated 256-bit keys stored with 0o600 permissions
✓ Lazy imports for optional dependencies (redis, cryptography) prevent errors in minimal environments
✓ PrivacyManager includes consent tracking, sensitive data classification, and audit logging
✓ SensitiveDataDetector classifies credentials and financial info as RESTRICTED (not stored)
✓ TokenBudgetManager uses Redis only; no credential exfiltration
✓ Redis adapter supports password authentication and TLS recommendations are documented
✓ All file paths are user-configurable with no hardcoded system paths
✓ Dependencies have minimum version constraints (pinned major versions)