The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: Provisions in the README.md template suggest installing the tool by piping a script from cloudweaver.fagnerlopes.dev (the author's domain) directly into a shell interpreter. This 'curl | bash' pattern is a high-risk vector for remote code execution.
[REMOTE_CODE_EXECUTION]: The hermes-host deployment workflow executes a remote installation script from hermes-agent.nousresearch.com using the curl | bash method during execution on the GitHub Actions runner.
[CREDENTIALS_UNSAFE]: The skill generates strong passwords and API keys for postgres and WAHA services, which are temporarily stored in plaintext in the user's home directory at ~/.cloud-weaver-${REPO_NAME}-report.json. While the skill intends to delete this file, the transient presence of plaintext secrets on disk is a credential exposure risk.
[CREDENTIALS_UNSAFE]: The setup process reads sensitive SSH private keys from ~/.ssh/ into shell variables to set them as GitHub Secrets, exposing potentially long-lived private key material in the session environment.
[COMMAND_EXECUTION]: The orchestration logic makes extensive use of the gh CLI and local shell utilities to manipulate SSH keys, manage git repositories, and configure infrastructure secrets.
[EXTERNAL_DOWNLOADS]: The AGENTS.md file contains directives to download and install additional skills via npx, which retrieves packages from the public npm registry.
[INDIRECT_PROMPT_INJECTION]: The skill includes 'bootstrap' files (AGENTS.md, CLAUDE.md) that use directional instructions to force the AI agent to load and follow a specific playbook before proceeding with user requests.

cloud-weaver-repo-setup