capability-evolver

SUSPICIOUS. The skill’s capabilities broadly match its stated purpose, but that purpose itself is a high-risk autonomous self-modification engine. Main concerns are immediate autonomous apply behavior, continuous looping, broad read/write scope, and provenance mismatch between the analyzed package identity and upstream official package branding/update docs. No clear evidence of outright credential theft or malicious exfiltration is shown from the provided content.

The code appears to implement an automated repository-evolution workflow, not overt malware. It has meaningful operational and supply-chain risks: an environment-provided shell command is executed directly, updates are forced automatically, and sensitive local context is placed in prompts/artifacts and may be sent to the hub. Review configuration and the omitted hub/executor modules before enabling it in sensitive environments.

This is a release/publishing automation script, not apparent malware. It performs expected remote publication actions and handles a GitHub token for release creation. The main security concern is command injection from untrusted environment/configuration values interpolated into execSync shell commands, plus the impact of its publishing and file-replacement operations. Use only trusted configuration and consider replacing shell command strings with argument-array subprocess calls and validating inputs.

This is intended artifact-cleanup code, not clear malware. It has a command-injection weakness because directory entry names are interpolated into shell commands; use an argument-based process API or direct unlink operations. Exploitability depends on an attacker being able to create crafted matching filenames in the evolution directory.