The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The file scripts/imgbb-upload.mjs contains a hardcoded, AES-encrypted API key (IMGBB_KEY_CIPHER_B64) that is decrypted at runtime using a derived key. This technique is used to embed a functional credential while evading simple static secret detection.
[DATA_EXFILTRATION]: The skill is configured to automatically upload local image files to the ImgBB service (api.imgbb.com) when performing image edits in the generations transport mode. This involves sending local filesystem data to a third-party server.
[COMMAND_EXECUTION]: The skill instructions in SKILL.md and README.md require the agent to execute shell commands (node scripts/gpt-image2.mjs, npm link, etc.) and manage local configuration files that store API keys in plain text.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection (Category 8) due to the following surface:
- Ingestion points: Untrusted data enters the agent context through the --image flag (processed files) and --var flag (user-provided template variables) in scripts/gpt-image2.mjs.
- Boundary markers: There are no explicit boundary markers or instructions to ignore embedded commands within the processed data.
- Capability inventory: The skill possesses network capabilities (fetch for API calls and ImgBB uploads) and file system access (fs.writeFileSync for saving images and fs.readFileSync for images/config).
- Sanitization: No sanitization or validation of the content of variables interpolated into prompts is performed before they are sent to the LLM-compatible image API.

gpt-image2