jira-expert
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill utilizes a custom Python script (
scripts/jql_query_builder.py) to map natural language descriptions to JQL queries. This process creates an attack surface where maliciously crafted input could potentially result in unintended JQL query execution through the agent's MCP tools.\n - Ingestion points: The natural language description passed as a command-line argument to the
scripts/jql_query_builder.pyscript (e.g., from user prompts or issue content).\n - Boundary markers: The instructions do not provide explicit boundary markers or directives for the agent to disregard potential instructions hidden within the data processed by the JQL builder script.\n
- Capability inventory: The skill uses Atlassian MCP tools (
mcp__atlassian__*) which allow for searching, creating, editing, and transitioning Jira issues.\n - Sanitization: The JQL builder script employs a pattern-matching and keyword-based approach, which serves as an implicit structural filter, and includes basic syntax checks for quotes and parentheses, but it lacks specific sanitization for prompt injection patterns targeted at manipulating the generated JQL logic.
Audit Metadata