manage-boards

Pass

Audited by Gen Agent Trust Hub on Aug 22, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes a bundled local Python script scripts/dashboard_utils.py to perform deterministic operations like title normalization, duplicate detection, and recency arithmetic.
  • [COMMAND_EXECUTION]: Instructions in SKILL.md direct the agent to suggest or execute the /mixpanel:install command to ensure the required execution environment (Mixpanel engine) is active.
  • [PROMPT_INJECTION]: The skill handles external data (dashboard titles, descriptions, and user-provided requirements) which presents a surface for indirect prompt injection. 1. Ingestion points: External dashboard metadata fetched via API and user-provided natural language descriptions in commands/create-dashboard.md. 2. Boundary markers: The instructions do not define specific delimiters for separating external strings from system prompts. 3. Capability inventory: The skill possesses capabilities for dashboard creation, layout modification, deletion, and query execution. 4. Sanitization: Title normalization logic in scripts/dashboard_utils.py and a text-card HTML whitelist described in references/mcp-tool-reference.md provide basic content filtering.
  • [SAFE]: The skill implements significant safety controls including mandatory user confirmation for all deletions and a "Validate every write" rule that performs an automated readback to verify that changes match the intended state.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 22, 2026, 01:46 PM