issue-triage
Pass
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data by reading GitHub issue descriptions and comments through
gh issue viewandWebFetch. This creates a surface for indirect prompt injection where malicious instructions embedded in an issue could attempt to manipulate the agent's behavior during analysis or response drafting.\n - Ingestion Points: Step 1 (获取 Issue 内容) utilizes
gh issue viewandWebFetchto ingest external content.\n - Boundary Markers: There are no explicit instructions or delimiters defined to isolate untrusted issue content from the agent's core instructions.\n
- Capability Inventory: The agent has the ability to read the local codebase for diagnosis (Step 2) and write back to GitHub using
gh issue comment(Step 6).\n - Sanitization: The process does not include explicit sanitization or filtering of the ingested issue content.\n
- Mitigation: The skill implements a robust human-in-the-loop workflow, requiring explicit user confirmation at every critical stage (Issue understanding, Qualitative assessment, Decision making, and Response posting), which significantly mitigates the risk of unauthorized actions.
Audit Metadata