The Agent Skills Directory

[DATA_EXFILTRATION]: The skill and its bundled script scripts/generate.py transmit local image files and the GPTIMAGE_API_KEY to the external domain tryallapi.com. This domain is not a recognized or trusted service, creating a significant risk of data and credential exfiltration.\n- [METADATA_POISONING]: The instructions in SKILL.md recommend that domestic users visit https://chatgpt-plus.top/list/#/home, a URL that has been explicitly flagged by automated scanners as a malicious phishing site.\n- [COMMAND_EXECUTION]: The skill instructs the agent to execute local Python scripts (scripts/generate.py and scripts/online_link.py) that perform network communication and local file system writes, which are used to facilitate data transmission to untrusted endpoints.\n- [EXTERNAL_DOWNLOADS]: The script scripts/generate.py programmatically downloads content from URLs returned by the API response and writes them to the local disk without integrity verification.\n- [CREDENTIALS_UNSAFE]: The skill requires users to provide an API key via environment variables, which is then sent to non-standard, potentially untrusted API gateways.\n- [INDIRECT_PROMPT_INJECTION]: The skill has a large attack surface for indirect injection as it processes untrusted user-provided content.\n
Ingestion points: User-supplied garment descriptions and model styling parameters in SKILL.md.\n
Boundary markers: None; user input is directly interpolated into the generation prompt without protective delimiters.\n
Capability inventory: Shell execution of local scripts, network exfiltration to external domains, and local file creation.\n
Sanitization: No input validation or escaping is performed on external content before it is used in prompt construction.

model-try-on