The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests and processes untrusted data from a SonarQube server, such as issue messages, security hotspot details, and user comments, which could potentially contain malicious instructions designed to influence the agent's behavior.
Ingestion points: Data is fetched from various SonarQube API endpoints in sonarqube-issues.sh, sonarqube-hotspots.sh, sonarqube-projects.sh, and sonarqube-api.sh.
Boundary markers: The skill does not implement explicit delimiters or instructions to ignore embedded commands when presenting API results to the agent.
Capability inventory: The skill provides several write capabilities, including transitioning issue statuses (resolve, confirm, false-positive) and adding comments via sonarqube-transition.sh, as well as raw API access via sonarqube-api.sh.
Sanitization: While the scripts use jq to parse JSON and urlencode for certain query parameters, the actual textual content from SonarQube (which could be influenced by an attacker who can commit code or add comments) is passed directly into the agent's context.
[COMMAND_EXECUTION]: The skill relies on shell scripts that execute CLI tools like curl, jq, and sed to perform its primary functions.
Evidence: The _api.sh helper script constructs and executes curl commands using the user-provided SONARQUBE_URL and SONARQUBE_TOKEN. The logic uses proper quoting to prevent local shell injection from agent-supplied arguments.

sonarqube