pp-surgegraph

Warn

Audited by Gen Agent Trust Hub on Jul 13, 2026

Risk Level: MEDIUMDATA_EXFILTRATIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSCREDENTIALS_UNSAFE
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill provides instructions to download and install its CLI tool using npx -y @mvanhorn/printing-press-library and go install github.com/mvanhorn/printing-press-library/.... These resources are hosted on public registries (NPM and GitHub) and belong to the skill's author, mvanhorn.
  • [DATA_EXFILTRATION]: The CLI includes a --deliver webhook:<url> parameter that enables the agent to POST command output directly to an external URL. This represents a functional data exfiltration path for sensitive project metrics, citation data, or generated content.
  • [COMMAND_EXECUTION]: The skill's primary operation involves executing shell commands via the surgegraph-pp-cli binary. It also uses npx for installation and doctor for environment verification.
  • [CREDENTIALS_UNSAFE]: The skill provides specific commands for managing sensitive credentials, such as create-api-key, update-api-key, and delete-api-key, supporting providers like OpenAI, Gemini, and Anthropic. While the CLI claims to mask keys in list outputs, it explicitly facilitates the storage and modification of these secrets.
  • [DATA_EXFILTRATION]: The --deliver file:<path> flag allows the agent to write command output to arbitrary local file paths. If misused, this could overwrite system configuration or shell profiles to achieve persistence or escalate privileges.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data that could contain malicious instructions.
  • Ingestion points: Fetches content from external WordPress sites via create-optimized-document, competitor topics via create-domain-research, and AI engine citation text via get-ai-visibility-citations.
  • Boundary markers: No explicit instruction delimiters or sanitization warnings are present in the CLI commands.
  • Capability inventory: The skill possesses the ability to execute shell commands, write to the file system, and perform network POST requests.
  • Sanitization: No evidence of sanitization or filtering of external content is provided before the data is processed or stored.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Jul 13, 2026, 05:42 PM
Security Audit — agent-trust-hub — pp-surgegraph