The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/quick-diagnostic.sh is vulnerable to shell command injection. It constructs a command string by appending the $SCHEME and $WORKSPACE variables (derived from user-provided command-line arguments) without sanitization, then executes the result using eval. An attacker providing a malicious scheme name containing shell metacharacters could achieve arbitrary command execution.\n- [DYNAMIC_EXECUTION]: The use of eval in scripts/quick-diagnostic.sh for executing dynamically assembled build commands constitutes a high-risk pattern. This enables arbitrary execution of shell commands if the input variables are manipulated.\n- [INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection due to its core workflow of processing external build logs and compiler output.\n
Ingestion points: Build output logs and error messages from xcodebuild and swiftc.\n
Boundary markers: Absent.\n
Capability inventory: Shell command execution (xcodebuild, swiftc, security), and file system modification (rm -rf).\n
Sanitization: Absent.\n- [DATA_EXFILTRATION]: The skill accesses sensitive developer-related file paths and system information. Specifically, it reads from ~/Library/MobileDevice/Provisioning Profiles and ~/Library/Developer/Xcode/DerivedData, and it executes the security command to retrieve code signing identities. While these actions are relevant to resolving iOS build and signing errors, they involve the exposure of sensitive environment metadata.

xcode-build-resolver