chongplus-image

Warn

Audited by Gen Agent Trust Hub on Oct 13, 2026

Risk Level: MEDIUMDATA_EXFILTRATIONPERSISTENCECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [DATA_EXFILTRATION]: The script scripts/chongplus_image.py reads a locally stored API key from the user's configuration directory and transmits it in the Authorization header to dynamic URLs provided by the remote service response in the download_image function. This pattern allows a potentially compromised or malicious server to exfiltrate the user's secret credential by returning an arbitrary third-party URL for image download.
  • [PERSISTENCE]: The SKILL.md instructions direct the agent to check for upstream updates once per day and suggest a workflow for authorizing and installing remote code updates from the Rodert/chongplus-image-skill repository. This establishes a mechanism for the skill to manage its own execution logic from a remote source, which could be exploited to introduce persistent malicious code.
  • [COMMAND_EXECUTION]: The skill requires the agent to execute a bundled Python client (chongplus_image.py) that interacts with the host filesystem to read environment variables, write configuration data, and manage output image files. The script uses the os, pathlib, and urllib modules for these operations.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes external data from the ChongPlus API and accepts local image files as input, creating a surface for indirect prompt injection if the agent interprets content from these sources as instructions.
  • Ingestion points: scripts/chongplus_image.py ingests user-supplied image files via the edit command and parses JSON-formatted responses from the ChongPlus API.
  • Boundary markers: No delimiters or safety instructions are defined to separate untrusted data from the agent's internal instruction stream during the processing of image metadata or API responses.
  • Capability inventory: scripts/chongplus_image.py includes capabilities for file modification (Path.write_bytes, os.replace) and network communication (urllib.request.urlopen).
  • Sanitization: The script does not perform content validation or sanitization on the image data or API responses beyond verifying JSON structure and Base64 encoding.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Oct 13, 2026, 09:33 AM