📦

Audit History

agentdb-learning-plugins - 10 audits

Version comparison

Capability and finding changes across audited versions, newest first.

VersionDateResultReview itemsChange vs previous
v10 Latest2026年7月9日 11:42 No confirmed findings0No capability change
v9 2026年7月9日 11:42 No confirmed findings0No capability change
v8 2026年7月5日 12:19 No confirmed findings9No capability change
v7 2026年7月5日 12:19 No confirmed findings9 Filesystem access
v6 2026年6月29日 18:14 No confirmed findings3Filesystem access
v5 2026年1月17日 04:17 No confirmed findings0No capability change
v4 2026年1月17日 04:17 No confirmed findings0Network accessExternal commands
v3 2026年1月10日 13:16 No confirmed findings0No capability change
v2 2026年1月10日 13:16 No confirmed findings0No capability change
v1 2026年1月10日 13:16 No confirmed findings0Baseline

2026年7月5日 12:19

Most static findings are Markdown code fences or illustrative TypeScript and JSON examples. The confirmed issue is repeated use of unpinned npx agentdb@latest commands, which can execute remote package code at runtime. No prompt injection, credential exfiltration, or malicious intent was found.

1
Files scanned
546
Lines analyzed
11
Review items
0
False positives ignored
Capability review items (9)

These are real local capabilities that may be expected for this skill, so they require review but are not counted as confirmed malicious behavior.

Medium
Ruby/shell backtick execution · 8 occurrences
```bash
The block contains several npx agentdb@latest commands that execute a package resolved at runtime. This is legitimate setup guidance, but it is an unpinned external command risk.
Medium
Ruby/shell backtick execution
```
The range includes an inline npx agentdb@latest mcp command in the learn-more section. It is documentation, but it can execute external package code if followed.
Audited by: codex

2026年7月5日 12:19

Most static findings are Markdown code fences or illustrative TypeScript and JSON examples. The confirmed issue is repeated use of unpinned npx agentdb@latest commands, which can execute remote package code at runtime. No prompt injection, credential exfiltration, or malicious intent was found.

1
Files scanned
546
Lines analyzed
11
Review items
0
False positives ignored
Capability review items (9)

These are real local capabilities that may be expected for this skill, so they require review but are not counted as confirmed malicious behavior.

Medium
Ruby/shell backtick execution · 8 occurrences
```bash
The block contains several npx agentdb@latest commands that execute a package resolved at runtime. This is legitimate setup guidance, but it is an unpinned external command risk.
Medium
Ruby/shell backtick execution
```
The range includes an inline npx agentdb@latest mcp command in the learn-more section. It is documentation, but it can execute external package code if followed.
Audited by: codex

2026年6月29日 18:14

The high static score is mostly caused by Markdown code fences and documented examples, not hidden executable code. The skill is publishable with a medium warning because it instructs users to run npx agentdb@latest and store learning data in a local database.

1
Files scanned
546
Lines analyzed
6
Review items
2
False positives ignored
Capability review items (3)

These are real local capabilities that may be expected for this skill, so they require review but are not counted as confirmed malicious behavior.

Medium
External Package Execution in Workflow Examples
The skill repeatedly instructs users to run npx agentdb@latest commands. This is legitimate AgentDB usage, but it executes a package fetched from the npm ecosystem and can create local files.
The commands are explicit and central to the workflow. They appear legitimate, but npx execution of a community package is a real supply-chain and local execution risk.
Medium
Local Learning Data Storage
The examples configure a local AgentDB database and insert agent experience data. This can store sensitive task data if users train on private observations or prompts.
The local database path and insertPattern examples are directly present. The sensitivity depends on user-provided training data, so the issue is contextual rather than malicious.
Low
Hardcoded Documentation URLs
The skill links to the AgentDB GitHub repository and website. These are documentation references and do not show data exfiltration or unauthorized network access.
The URLs are visible Learn More links. No evidence shows credentials, local data, or user content being sent to these endpoints.
Static false positives ignored (2)

These static matches were dismissed by semantic review or matched schema-only tokens, so they are shown for transparency but do not drive the quality score.

Low
Markdown Backtick False Positives
Most Ruby backtick execution alerts are Markdown code fences around bash, TypeScript, and JSON examples. They are documentation blocks, not executable skill code.
The surrounding lines show fenced documentation examples. There is no Ruby source file, shell wrapper, eval behavior, or hidden execution path in the skill.
Low
Weak Cryptography Alerts Are Unsupported
The weak cryptography alerts point to description and example lines that do not contain hash functions or cryptographic algorithms. No evidence found for weak crypto usage.
The cited lines discuss AgentDB capabilities and training loops, not MD5, SHA1, DES, RC4, or similar weak algorithms. This is a high-confidence false positive.

Detected Patterns

Supply-Chain Sensitive npx Usage
Audited by: codex

2026年1月10日 13:16

Documentation-only skill containing no executable code. Provides guidance for using AgentDB's external CLI tool and API. No network calls, file system access, or code execution paths defined in the skill file itself.

1
Files scanned
546
Lines analyzed
0
Review items
0
False positives ignored
No confirmed security findings were recorded for this completed audit.
Audited by: claude

2026年1月10日 13:16

Documentation-only skill containing no executable code. Provides guidance for using AgentDB's external CLI tool and API. No network calls, file system access, or code execution paths defined in the skill file itself.

1
Files scanned
546
Lines analyzed
0
Review items
0
False positives ignored
No confirmed security findings were recorded for this completed audit.
Audited by: claude

2026年1月10日 13:16

Documentation-only skill containing no executable code. Provides guidance for using AgentDB's external CLI tool and API. No network calls, file system access, or code execution paths defined in the skill file itself.

1
Files scanned
546
Lines analyzed
0
Review items
0
False positives ignored
No confirmed security findings were recorded for this completed audit.
Audited by: claude