graphify

SUSPICIOUS: the skill’s core purpose and dependency provenance are mostly coherent, with official package-manager installation and proportionate Gemini credential use. The main risk is its large automation footprint: it processes untrusted external content and explicitly dispatches write/bash-capable subagents, creating a notable indirect prompt-injection and execution-risk surface even without clear malicious intent.

No clear evidence of deliberate malware (e.g., credential theft, backdoor, or exfiltration) appears in the provided snippet. The primary security concern is a high-impact local tampering control point: the tool executes a command/interpreter determined by the contents of graphify-out/.graphify_python via shell command substitution. Separately, the orchestration fetches and processes arbitrary user-provided URLs and can continuously process untrusted watched files, increasing exposure to downstream toolchain vulnerabilities or unsafe input handling. Review integrity/permissions for graphify-out/.graphify_python and inspect graphify.ingest/graphify.watch for URL/file validation and safe network/file handling.