wordcraft
Pass
Audited by Gen Agent Trust Hub on Nov 16, 2026
Risk Level: SAFECREDENTIALS_UNSAFECOMMAND_EXECUTIONDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [CREDENTIALS_UNSAFE]: The skill instructions direct the agent to read an authentication key from a local file (e.g.,
~/Library/Application Support/WordCraft/control-key.local-7981). This key grants the agent the ability to authenticate and control the WordCraft application session. - [COMMAND_EXECUTION]: The agent is empowered to execute arbitrary automation commands within the WordCraft environment via the
engine.executemethod. This includes document manipulation, style modification, and simulated UI interactions likeui.clickandui.key. - [DATA_EXFILTRATION]: The skill provides tools for capturing the document's content and appearance through
document.inspect,ui.screenshot(capturing the window state), andui.render(rendering pages to PNG). This presents a data exfiltration surface if sensitive document content is sent to external services. - [INDIRECT_PROMPT_INJECTION]: The skill processes document content which may originate from untrusted sources, creating a potential vulnerability to indirect prompt injection.
- Ingestion points: Document blocks, styles, and runs are read via
document.inspectand processed by the agent. - Boundary markers: No explicit delimiters or instructions to ignore embedded commands are specified for the ingested document content.
- Capability inventory: The skill possesses significant capabilities including command execution (
engine.execute), UI automation (ui.click,ui.key), and visual capture (ui.screenshot,ui.render). - Sanitization: There are no instructions for sanitizing or escaping the content read from documents before it is processed by the AI.
Audit Metadata