day2-create-context-sync-skill

Pass

Audited by Gen Agent Trust Hub on Sep 22, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSDYNAMIC_EXECUTIONPERSISTENCEINDIRECT_PROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
  • [COMMAND_EXECUTION]: The skill instructions direct the agent to execute installation commands such as npx skills add ai-native-camp/camp-1 and use uv run python to execute local scripts.
  • [EXTERNAL_DOWNLOADS]: The skill workflow begins by downloading and installing additional skill assets from the vendor's repository (ai-native-camp/camp-1) using the npx skills utility. It also references the installation of the @anthropic-ai/mcp-fetch tool from a well-known service.
  • [DYNAMIC_EXECUTION]: The skill architecture relies on the agent dynamically generating Python scripts at runtime (e.g., gmail_fetch.py, calendar_fetch.py) to interface with external APIs, which are then executed within the local environment.
  • [PERSISTENCE]: The skill includes explicit instructions for establishing persistent execution by adding automated schedules to the CLAUDE.md file (e.g., daily runs at 09:00), allowing the synchronization logic to run without direct user initiation.
  • [INDIRECT_PROMPT_INJECTION]: The skill creates a vulnerability surface by ingesting and processing untrusted data from multiple external sources like Slack, Gmail, and Notion.
  • Ingestion points: Content is retrieved from Slack channels, Gmail inboxes, Notion databases, Linear issues, and GitHub repositories as defined in references/block0-tool-selection.md and templates/context-sync.md.
  • Boundary markers: The provided templates and instructions do not include specific delimiters or warnings to prevent the agent from interpreting instructions that may be embedded within the processed messages or emails.
  • Capability inventory: The skill environment allows for file system writes, command-line execution (npx, python), and network access through various Model Context Protocol (MCP) toolsets.
  • Sanitization: There is no evidence of content sanitization or validation logic to filter out potentially malicious instructions from the ingested external data.
  • [DATA_EXFILTRATION]: The skill's primary function is to read sensitive user data from private communication and management tools to consolidate it into reports, which can then be transmitted to external platforms like Slack or Notion.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 22, 2026, 09:31 AM