synthesis-mac-sync

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The skill instructs the assistant to read and sync user config files (including files marked as containing API keys), to display diffs and manifest remote URLs during reconciliation, and to echo manifest_url/existing_url in summaries—actions that can cause secret values (API keys, tokens embedded in files or URLs) to be included verbatim in outputs.

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.90). The skill's workflow explicitly fetches and pulls from git remotes (e.g., GitHub) during the "Git repo sync" and reads/writes config files from the user's iCloud sync folder and workspace repos (including .agents files), which are untrusted, user-generated third-party sources whose contents the agent ingests and can change behavior (pulling code, creating symlinks, or executing one-time actions).

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.80). This skill performs runtime git fetch/pull using repository URLs (e.g., https://github.com/user/my-app.git) to update workspace-private repos whose .agents/*.yaml files are symlinked and consumed by agent tools, so fetched remote repo content can directly control agent prompts/configuration.

MEDIUM W013: Attempt to modify system services in skill instructions.

• Attempt to modify system services in skill instructions detected (high risk: 0.90). The skill instructs the agent to run batched shell scripts that modify many local configuration files (including ~/.ssh/config and other sensitive dotfiles), add/update git remotes, create/remove symlinks, change permissions, and push/pull repos automatically—actions that modify the machine's state and can be destructive even without sudo.