The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructions direct the agent to execute a local Node.js script via the shell to perform astrological calculations.
Evidence: SKILL.md Step 3 provides explicit shell commands for the agent to run, such as node "<absolute-bazi-skill-directory>/scripts/calculate.js".
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted user data (birth date, time, and location) which is then used as input for a command-line script.
Ingestion points: User birth details (birthDate, birthTime, longitude, etc.) defined in SKILL.md and processed in lib/analyze.js.
Boundary markers: The skill lacks explicit instructions to sanitize or escape user input beyond formatting it as a JSON string.
Capability inventory: The skill has the capability to execute shell commands using the node binary.
Sanitization: No specific sanitization or filtering logic is provided in the instruction files for the user-supplied metadata.
[DYNAMIC_EXECUTION]: The skill relies on executing its own JavaScript files at runtime to generate findings, which is a form of local script execution.
Evidence: The skill defines core logic in lib/analyze.js and scripts/calculate.js which are executed by the environment via Node.js.

bazi