daily-transit

Pass

Audited by Gen Agent Trust Hub on Jul 29, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests data from a shared evaluator tool, which presents an attack surface for indirect prompt injection if the tool output is manipulated by an external actor.
  • Ingestion points: The skill processes output from the command shared evaluator --topic daily-transit mentioned in SKILL.md.
  • Boundary markers: The instructions include explicit constraints: 'Use its resolved timestamp and cited transit facts' and 'Do not add uncited mood, event, or timing claims.'
  • Capability inventory: The skill code consists only of instructions and references to other local documentation (AGENTS.md, SKILL.md); it does not contain network operations or file-system modifications.
  • Sanitization: The prompt provides clear guidelines for the agent to stick strictly to cited evidence, effectively limiting the influence of potentially injected data.
  • [COMMAND_EXECUTION]: The skill requires the agent to run a command-line tool to retrieve data.
  • Evidence: The instruction in SKILL.md to 'run the shared evaluator with --topic daily-transit' is a functional requirement to perform its stated purpose.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 29, 2026, 05:41 PM
Security Audit — agent-trust-hub — daily-transit