The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data retrieved from Gitee issue titles and descriptions (via the list_repo_issues and get_repo_issue_detail tools). This content is used to inform the agent's triage decisions, including priority assessment and label assignment.
Ingestion points: Untrusted data enters the agent context from the output of Gitee MCP tools in Step 1 and Step 2.
Boundary markers: The instructions do not specify any delimiters or specific boundary markers to isolate the untrusted issue content from the agent's instructions.
Capability inventory: The skill possesses write capabilities through the update_issue and create_comment tools.
Sanitization: There are no explicit instructions for the agent to sanitize or ignore embedded instructions within the processed issues.
Mitigation: A significant security control is present in Step 4, which requires explicit user confirmation before any bulk updates or comments are performed, providing a human-in-the-loop check against potential injection attacks.

triage-issues