roast-my-product

Pass

Audited by Gen Agent Trust Hub on Aug 1, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill instructs the agent to execute bash scripts during its preamble and completion phases. These scripts are used for session management and telemetry logging.\n- [DATA_EXPOSURE]: The agent reads from ~/.superstack/config.json and writes to ~/.superstack/telemetry.jsonl to manage user preferences and log telemetry events.\n- [EXTERNAL_DOWNLOADS]: The skill uses curl to send usage statistics to a remote server (Convex). While this is a common telemetry pattern, it involves outbound network communication. The skill appropriately uses a consent prompt before enabling this feature.\n- [PROMPT_INJECTION]: The skill processes untrusted data from local files (.superstack/idea-context.md and .superstack/build-context.md) to provide its critique. This creates a surface for indirect prompt injection as there are no boundary markers or sanitization steps described for this input.\n
  • Ingestion points: Reads product context from .superstack/idea-context.md and .superstack/build-context.md in Step 2 of the workflow.\n
  • Boundary markers: None; the content is interpolated directly into the agent's reasoning process without delimiters or instructions to ignore embedded commands.\n
  • Capability inventory: The skill has the ability to execute shell commands (bash) and interact with the user via AskUserQuestion.\n
  • Sanitization: None; the ingested text is not filtered or escaped before processing.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 1, 2026, 11:05 AM
Security Audit — agent-trust-hub — roast-my-product