productivity-start
Pass
Audited by Gen Agent Trust Hub on Apr 18, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill instructions in Steps 5 and 6 guide the agent to ingest and analyze data from potentially untrusted sources such as emails, messages, and external task lists (Jira, Notion, Asana). This creates a surface for Indirect Prompt Injection.
- Ingestion points: Data enters the agent context from user-provided task files, external productivity applications, chat logs, emails, documents, and calendars (SKILL.md).
- Boundary markers: The skill fails to provide delimiters or specific instructions to the agent to ignore or isolate commands that may be embedded within the ingested data.
- Capability inventory: The skill checks the local file system and creates local task and memory files (TASKS.md, CONTEXT.md) but does not include direct network exfiltration or shell execution logic in this file.
- Sanitization: No procedures for data validation, escaping, or filtering of the external data sources are defined before the content is processed by the AI.
Audit Metadata