The Agent Skills Directory

[COMMAND_EXECUTION]: Documentation includes a sudo command (sudo ufw allow 4433/udp) which requires root privileges to modify system firewall rules for QUIC synchronization.
[REMOTE_CODE_EXECUTION]: Instruction to use npx agentdb@latest which downloads and executes the latest version of a package from the npm registry, representing a remote code execution risk.
[EXTERNAL_DOWNLOADS]: The skill relies on external Node.js packages agentic-flow and agentdb from public registries.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface.
Ingestion points: The insertPattern function accepts external data such as documentText in SKILL.md.
Boundary markers: No delimiters or ignore instructions are present in the provided examples.
Capability inventory: The skill performs subprocess calls via CLI and network operations via the QUIC protocol.
Sanitization: No sanitization or validation of ingested text is performed before it is synthesized into context.
[PROMPT_INJECTION]: Metadata analysis reveals a potential character-substitution pattern between the author handle aegntic and the referenced toolset agentic-flow (ae vs ag), which can be a sign of deceptive naming.

AgentDB Advanced Features