ios-auto-release
Fail
Audited by Gen Agent Trust Hub on Sep 4, 2026
Risk Level: HIGHPERSISTENCECREDENTIALS_UNSAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONDYNAMIC_EXECUTION
Full Analysis
- [PERSISTENCE]: The skill provides a template (
templates/device-install.plist) and instructions to install it as a macOS LaunchAgent in~/Library/LaunchAgents/. This creates a persistent background process (watch_release_install.sh) that runs at every login and monitors the repository for changes to trigger automated builds.- [CREDENTIALS_UNSAFE]: The documentation (SKILL.md) contains a hardcoded Apple Developer Team ID (GU9WTSTX9M). It also instructs users to store sensitive signing credentials, including App Store Connect API keys and paths to private.p8files, in a localorg-credentials.envfile. This practice increases the risk of local credential exposure and unintentional inclusion in backups or other files.- [INDIRECT_PROMPT_INJECTION]: Thescripts/ci_submit.pyscript ingests data from aCHANGELOG.mdfile located in the target repository to generate 'What's New' release notes for App Store Connect. 1. Ingestion points:CHANGELOG.mdfile read byci_submit.py. 2. Boundary markers: None; the script extracts text based on regex headers. 3. Capability inventory: The script has capabilities to POST and PATCH data to the App Store Connect API. 4. Sanitization: The script performs minimal cleaning, removing only markdown formatting characters like '*' and '' from the extracted text.- **[DYNAMIC_EXECUTION]:** Thescripts/install_on_device.shscript executes an embedded Python script using a heredoc pattern (python3 - "$WANT" <<'PY'
). It also usesevalto resolve the path for theASC_PRIVATE_KEY_PATHenvironment variable, which presents a command injection risk if environment variables are manipulated.- **[COMMAND_EXECUTION]:** The skill relies on various shell commands (xcodebuild,security,xcrun`) within its scripts and GitHub Actions templates to perform archiving, signing, and uploading tasks using environment-derived variables.
Recommendations
- AI detected serious security threats
Audit Metadata