The Agent Skills Directory

[Indirect Prompt Injection]: The skill is designed to ingest and process untrusted data from GitHub issues, including free-form text and code snippets provided by users. This creates a surface where malicious instructions embedded in an issue could influence the agent's behavior during the triage process.
Ingestion points: SKILL.md (Step 1) instructs the agent to read the issue and linked resources.
Boundary markers: None specified for delimiting user-provided content from instructions.
Capability inventory: SKILL.md (Step 5.2) requires the agent to compile and execute a console program based on user input.
Sanitization: The skill includes a proactive safety check (Step 2) instructing the agent to exit immediately if a security concern is suspected.
[Execution of Generated Code]: The skill requires the agent to generate, compile, and execute a C# console application to reproduce reported bugs. Because this code is partially derived from user-submitted snippets, there is a possibility that malicious code could be inadvertently executed in the triage environment.
[Potential Data Exposure]: The provided code template utilizes an environment variable (Test__SqlServer__DefaultConnection) for database connectivity. Maliciously crafted issue reports could attempt to include code that prints or exfiltrates the value of this environment variable or other system information when the agent runs the reproduction script.
[Command Execution]: The skill uses the GitHub CLI (gh) to interact with the repository and manage labels. While these are legitimate triage operations, they represent an automated interaction with an external platform based on analyzed content.

triage