assessing-jira-issue-relevance

Pass

Audited by Gen Agent Trust Hub on Aug 4, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection because it retrieves and interprets untrusted text from Jira ticket descriptions and comments to drive its search logic and tool usage. An attacker-controlled ticket could contain instructions designed to manipulate the agent's behavior.\n
  • Ingestion points: SKILL.md uses the tools get_issue and get_issue_comments to fetch external content from Jira.\n
  • Boundary markers: Absent. Content from the tickets is not wrapped in delimiters or provided with warnings to ignore embedded instructions.\n
  • Capability inventory: The skill has access to powerful tools including git clone, git log, and file reading capabilities.\n
  • Sanitization: Absent.\n- [COMMAND_EXECUTION]: The skill utilizes Bash(git clone:*) to download source code repositories. While the instructions include a requirement for user confirmation via AskUserQuestion before cloning, the repository target is determined by data fetched from external Jira tickets, creating a potential path for downloading untrusted content into the environment.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 4, 2026, 01:13 AM
Security Audit — agent-trust-hub — assessing-jira-issue-relevance