The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the Bash tool to execute curl commands for interacting with the SonarCloud REST API. While these commands use user-provided project keys and branch names as arguments, the primary risk is controlled within the scope of intended functionality.
[DATA_EXFILTRATION]: The skill requires and accesses the SONARCLOUD_TOKEN environment variable. It performs network requests (curl) to https://sonarcloud.io. Because SonarCloud is a well-known technology service and the skill follows standard practices (using environment variables rather than hardcoding), this is documented as a functional requirement rather than a malicious pattern.
[INDIRECT_PROMPT_INJECTION]: The skill reads and processes external data from SonarCloud (issue messages, file paths, and metrics).
Ingestion points: Data enters the context via curl responses from sonarcloud.io/api/* endpoints.
Boundary markers: None explicitly defined in the instructions to separate API data from agent instructions.
Capability inventory: The skill has Bash access, allowing it to execute shell commands based on the findings.
Sanitization: No specific sanitization or escaping of the API response content is mentioned before presenting it to the agent/user. An attacker with control over SonarCloud project metadata or code comments could potentially inject instructions into the analysis reports.

sonarcloud