Agent Designer
v2.1.1Agent Designer - Multi-Agent System Architecture
⭐ 0· 942·6 current·6 all-time
byAlireza Rezvani@alirezarezvani
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
OpenClaw
Benign
medium confidencePurpose & Capability
The name/description (agent architecture, tool/schema generation, evaluation) align with the included files (agent_planner.py, tool_schema_generator.py, agent_evaluator.py) and provided samples. Minor inconsistency: registry metadata declares no required binaries, but the package provides Python scripts that the README shows running via 'python ...'. A legitimate user-run workflow would require Python (and possibly standard Python libraries).
Instruction Scope
The SKILL.md and README describe design and evaluation workflows (generate architecture JSON/diagram, create tool schemas, analyze logs). The instructions do not direct the agent to read unrelated system secrets, modify other skills, or exfiltrate data. They reference reading/writing JSON and sample assets (expected for this purpose).
Install Mechanism
There is no external install spec or network download — code is shipped with the skill. No installers or archive downloads are present, which reduces supply‑chain risk compared to fetching remote executables.
Credentials
The skill declares no required environment variables or credentials, which is proportionate for an offline design/evaluation toolkit. Note: it uses/produces tool schemas for external integrations (OpenAI/Anthropic formats) but does not declare API keys — that is appropriate if the code only emits schemas. The omission of a declared 'python' binary is a small mismatch.
Persistence & Privilege
The skill is not always-enabled, does not request elevated privileges, and does not claim to change other skill configs. It reads sample input files and writes output artifacts (JSON, diagrams) — appropriate for its purpose.
Assessment
This package appears coherent with its stated purpose, but exercise normal caution before running code: 1) Ensure you have Python installed (the README demonstrates running python scripts) and run the scripts in a controlled environment (sandbox, VM, or container) if you cannot review code fully. 2) Inspect the three Python files for any network or subprocess behavior you don't expect (search for uses of 'requests', 'urllib', 'socket', 'subprocess', 'os.environ', or direct file writes outside working directories). 3) If you plan to integrate generated schemas with live services, provide API keys only to the components that actually need them and follow least‑privilege practices. 4) Confirm output paths when running the scripts so they do not overwrite important system files. If you want, I can scan the full sources for network calls, subprocess usage, or references to environment variables/secret access and report any risky lines to help you decide.Like a lobster shell, security has layers — review code before you run it.
latestvk97fq06sbgp8temmcbad0zsj0982mmh2
License
MIT-0
Free to use, modify, and redistribute. No attribution required.