The Agent Skills Directory

[UNVERIFIABLE_DEPENDENCIES_AND_REMOTE_CODE_EXECUTION]: The skill requires the installation of a third-party command-line interface via Homebrew (brew install ossianhempel/tap/things3-cli). This external dependency is not from a verified or well-known vendor.
[DYNAMIC_EXECUTION]: Structural edits such as renaming or deleting areas/projects are performed via osascript (AppleScript) commands built dynamically at runtime. This execution method poses a risk if user-controlled input, like task or project titles, is interpolated into the script without thorough sanitization.
[CREDENTIALS_UNSAFE]: The skill uses a THINGS_AUTH_TOKEN for modifying data. While the instructions suggest using environment variables or a local things.local.md file, the handling of this sensitive credential requires careful management to prevent exposure in logs or terminal history.
[INDIRECT_PROMPT_INJECTION]: The skill processes task data retrieved from the local Things 3 SQLite database. If tasks are created from untrusted sources, such as via the email-to-Things feature, the content could contain malicious instructions. The presence of powerful capabilities like AppleScript execution increases the impact of such an injection.
Ingestion points: Local Things 3 SQLite database queried via the things CLI.
Boundary markers: None identified for delimiting or neutralizing data retrieved from the database.
Capability inventory: things CLI (subprocess calls), osascript (shell command execution), and the things:/// URL scheme.
Sanitization: No explicit sanitization, escaping, or filtering logic is described for data interpolated into shell commands, AppleScript strings, or URLs.

things-app