daily-meeting-update
Fail
Audited by Gen Agent Trust Hub on Feb 16, 2026
Risk Level: HIGHPROMPT_INJECTIONCOMMAND_EXECUTIONDATA_EXFILTRATION
Full Analysis
- [Indirect Prompt Injection] (HIGH): The skill processes untrusted data from external sources (GitHub PRs, Jira tickets) while possessing command execution capabilities. \n
- Ingestion points: Output from
ghandjiraCLI tools. \n - Capability inventory: Executes
gh,git, andjiracommands; reads local files. \n - Boundary markers: None mentioned. \n
- Sanitization: None mentioned. \n- [Data Exposure] (HIGH): The skill specifically accesses the
~/.claude/projectsdirectory. This contains.jsonlfiles of Claude session history, which may include proprietary code, logic, and inadvertently exposed credentials or sensitive context. \n- [Command Execution] (MEDIUM): The skill relies on executing system commands (gh,git,jira) to function. While intended for data retrieval, these tools can be misused if the agent is manipulated via injection.
Recommendations
- AI detected serious security threats
Audit Metadata