skills/openshift-hyperfleet/hyperfleet-claude-plugins/JIRA Story Point Estimator/Gen Agent Trust Hub
JIRA Story Point Estimator
Fail
Audited by Gen Agent Trust Hub on Feb 16, 2026
Risk Level: HIGHPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- Indirect Prompt Injection (HIGH): The skill lacks sufficient isolation when processing untrusted content from JIRA tickets. An attacker who can edit a JIRA ticket could inject instructions to override the agent's logic.\n
- Ingestion Point: The skill uses
jira issue view TICKET-KEYto fetch ticket details, including descriptions and acceptance criteria, which are external, untrusted data sources.\n - Boundary Markers: No delimiters or isolation markers are defined in the instructions to separate untrusted ticket content from the agent's system instructions.\n
- Capability Inventory: The skill contains commands to modify the JIRA database, specifically
jira issue edit, which creates a high-risk path for an injected instruction to perform unauthorized modifications.\n - Sanitization: There is no evidence of input validation or sanitization before the retrieved data is processed by the AI for estimation analysis.\n- Command Execution (MEDIUM): The skill executes shell commands using the
jiraCLI. While intended for its core functionality, the combination of executing commands based on untrusted input without strict validation of the input fields poses a risk of argument injection or unintended tool usage.
Recommendations
- AI detected serious security threats
Audit Metadata