skills/sharpdeveye/maestro/capture/Gen Agent Trust Hub

capture

Warn

Audited by Gen Agent Trust Hub on Sep 14, 2026

Risk Level: MEDIUMDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [DYNAMIC_EXECUTION]: The skill constructs file paths using a user-controllable 'topic' argument to save session records (e.g., .maestro/sessions/{date}_{topic}.md). This pattern can lead to path traversal vulnerabilities if the user-provided topic contains characters like ../, potentially allowing the agent to write files outside of the intended directory.
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to read and summarize the current conversation history, which constitutes untrusted input.
  • Ingestion points: The conversation history is read globally to identify commands, decisions, and changes.
  • Boundary markers: The instructions do not define clear boundaries or specify that the agent should ignore instructions embedded within the conversation history being summarized.
  • Capability inventory: The skill possesses the capability to write new files and append to existing JSONL files in the .maestro/ directory.
  • Sanitization: There is no explicit requirement for the agent to sanitize or escape the content extracted from the conversation history before writing it to the filesystem.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Sep 14, 2026, 08:01 PM
Security Audit — agent-trust-hub — capture