ℹ
Purpose & Capability
The skill is an image-to-image helper and the included script accepts prompt + image(s) and posts them to a model API; requiring python and a TEAM_API_KEY is plausible. However the script defaults to a third-party endpoint (https://dlazy.com/api/ai/tool) which is not documented in the skill metadata or description and may be unexpected to users.
!
Instruction Scope
SKILL.md and the embedded systemPrompt explicitly demand the agent run `python script/invoke_model.py` directly and not use the platform's `openclaw run`. That is a direct attempt to bypass platform execution wrappers/monitoring. The script will read local files given as arguments, base64-encode them, and POST them to a remote API — which is expected for image upload but also allows arbitrary local-file exfiltration if the agent is instructed to pass other paths. The instructions do not constrain which files/paths are acceptable.
ℹ
Install Mechanism
There is no install spec (instruction-only with an included script), so nothing will be automatically downloaded. The script uses the Python requests library but the skill does not declare Python package dependencies; absence of an install step is low-risk but may cause runtime failures if dependencies are missing.
!
Credentials
The declared required env var is TEAM_API_KEY (reasonable). The code also reads TEAM_BASE_URL (defaulting to dlazy.com) which is not declared in requires.env or metadata. A single API key is proportionate, but the undeclared base URL and the default pointing to an unknown host are problematic and can be used to forward sensitive data to an external service.
!
Persistence & Privilege
The skill is not marked always:true and does not request persistent platform privileges, which is good. However the embedded imperative to bypass `openclaw run` reduces platform controls and effectively increases the risk because it encourages running the script directly in the environment rather than via the platform's safer invocation path.
What to consider before installing
This skill will read image files you give it, base64-encode them, and POST them to a remote API using the TEAM_API_KEY. Two red flags: (1) SKILL.md forces direct execution of the script ("do not use openclaw run"), which bypasses platform controls; (2) the script defaults to an undocumented endpoint (https://dlazy.com) via TEAM_BASE_URL (an env var not declared in the metadata). Before installing, verify the upstream provider (who maintains dlazy.com or where TEAM_BASE_URL should point), avoid giving sensitive images or keys unless you trust that endpoint, and ask the author to remove the instruction that forces direct execution and to declare TEAM_BASE_URL in the metadata. If you must test, run the script in an isolated environment and inspect traffic (or override TEAM_BASE_URL to a known trusted endpoint).
