The Agent Skills Directory

[CREDENTIALS_UNSAFE] (HIGH): Hardcoded database credentials detected in code samples.
Evidence: references/patterns.md contains a connection string Host=localhost;Database=sorcha_wallet;Username=postgres;Password=postgres. While these are likely local defaults, the inclusion of literal credentials instead of placeholders violates security best practices and exposes the database to unauthorized access if used in real environments.
[PROMPT_INJECTION] (HIGH): High-risk Indirect Prompt Injection vulnerability surface (Category 8c).
Ingestion points: The skill uses the mcp__context7__query-docs tool in SKILL.md to fetch external content from the library /dotnet/entityframework.docs.
Boundary markers: Absent. No instructions are provided to the agent to treat the external documentation as untrusted or to delimit it from its internal instructions.
Capability inventory: The skill explicitly allows the Bash, Write, and Edit tools in SKILL.md.
Sanitization: Absent. There is no verification of the documentation source or sanitization of its content before processing.
Analysis: This combination allows for a 'remote instruction execution' scenario where a malicious actor influencing the documentation can trick the agent into running dangerous shell commands.

entity-framework