The Agent Skills Directory

Indirect Prompt Injection Surface: The skill is designed to process untrusted data from external issue trackers and emails, which presents a potential surface for indirect prompt injection.\n
Ingestion points: Data enters the agent's context through gh issue view (capturing issue bodies and comments) and Gmail thread retrieval.\n
Boundary markers: The skill explicitly instructs the agent to treat external content as data rather than instructions and uses a "redact-after-fetch" protocol to mitigate risks.\n
Capability inventory: The skill has access to shell execution via the gh CLI and network-based tools via Gmail MCP.\n
Sanitization: To protect against command injection, the skill uses temporary files for CLI input and regex-based validation for specific identifiers like CVE IDs.\n- Command Execution: The skill performs command execution for operational purposes, including the use of the gh CLI and the execution of local Python-based pre-flight and privacy checks. These commands are localized to the project environment and are used for administrative tasks.

issue-triage