systemic-issue-triage
Warn
Audited by Gen Agent Trust Hub on Oct 5, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to verify suggested fixes or 'runnable continuations' by executing the specific command provided within the issue report. This is a direct vector for command injection, as an attacker could include malicious shell commands in a bug report that the agent would then execute during the triage process.
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external content from bug reports and community feedback to drive internal logic and tool usage.
- Ingestion points: Bug reports, community reports, and issue backlog items.
- Boundary markers: None identified; instructions do not require the agent to isolate or sanitize external text.
- Capability inventory: The agent is granted capabilities for repository-wide searching (grep), binary reproduction, and command execution.
- Sanitization: None identified; the agent is encouraged to 'trust reproduction over the issue text' but is not warned to sanitize the input that generates that reproduction.
- [DYNAMIC_EXECUTION]: The instructions require the agent to reproduce bug reports on 'fresh binaries' and execute 'printed commands' from reports. This involves running code or binaries whose source is derived from untrusted external descriptions, potentially leading to the execution of malicious payloads.
Audit Metadata