The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the agent to install binaries from the vendor's GitHub repository and npm registry using go install and npx. While these are vendor-owned resources, they involve downloading and executing external code during setup.
[DATA_EXFILTRATION]: The CLI tool supports a --deliver webhook:<url> flag, which allows the agent to POST command outputs to arbitrary external URLs. This represents a significant vector for data exfiltration if the agent is manipulated into sending sensitive information.
[DATA_EXFILTRATION]: The feedback command can be configured via environment variables (NVD_FEEDBACK_ENDPOINT and NVD_FEEDBACK_AUTO_SEND) to automatically transmit local data from ~/.nvd-pp-cli/feedback.jsonl to a remote server.
[COMMAND_EXECUTION]: The skill utilizes the nvd-pp-cli binary to perform searches and management tasks. It includes a --deliver file:<path> capability that allows the agent to write output to arbitrary file paths, potentially leading to unauthorized file creation or overwriting of system files.
[INDIRECT_PROMPT_INJECTION]: The skill ingests and processes vulnerability data (CVEs, CPEs) from the National Vulnerability Database, which is an external, untrusted source.
Ingestion points: Results from search-cves and search-cpes commands in SKILL.md.
Boundary markers: No specific delimiters or warnings are provided to the agent to ignore instructions embedded within the vulnerability data.
Capability inventory: The tool possesses the ability to write to the file system (--deliver file:) and perform network operations (--deliver webhook:).
Sanitization: There is no evidence of sanitization or filtering of the NVD content before it is processed by the agent.

pp-nvd