The Agent Skills Directory

[COMMAND_EXECUTION]: The skill employs a shell script (scripts/pilot.sh) to manage API interactions. It uses curl to send requests to https://staging.kalodata.com and utilizes process IDs (PIDs) to track background execution and polling status.
[DATA_EXFILTRATION]: User authentication is handled through a KaloData API token stored in the user's home directory at ~/.kalopilot/token. This token is transmitted to the official service domain (kalodata.com) for authentication. This is an expected pattern for skills integrating with third-party APIs.
[COMMAND_EXECUTION]: The scripts/pilot.sh script performs basic string interpolation to build JSON payloads (e.g., payload="{\"query\": \"$question\"}"). While this does not present a shell injection risk when variables are handled by the agent, it represents a best-practice violation as it could lead to invalid JSON if user input contains unescaped double quotes.
[PROMPT_INJECTION]: The skill has a surface for indirect prompt injection as it processes and displays data fetched from an external API (staging.kalodata.com). The agent is instructed to render text and report fields from the API response directly into the conversation.
Ingestion points: Data is received via the scripts/pilot.sh result command which reads from ~/.kalopilot/result.json.
Boundary markers: No specific delimiters or safety instructions are provided to the agent to treat the API response as untrusted data.
Capability inventory: The skill has capabilities to execute shell commands via scripts/pilot.sh (including curl and rm) and read/write files in the ~/.kalopilot/ directory.
Sanitization: No sanitization or validation of the API-returned content is performed before presentation to the user.

kalopilot