The Agent Skills Directory

[DYNAMIC_EXECUTION]: The helper script scripts/lib/background-scan.mjs uses the Function constructor to evaluate numeric literals and arithmetic expressions encountered during code scanning.
Evidence: The evaluateNumber function in scripts/lib/background-scan.mjs executes code via Function to parse intervals and budget constants.
Mitigation: The script applies a strict character filter that restricts input to digits, underscores, spaces, and basic arithmetic operators, preventing arbitrary code execution.
[INDIRECT_PROMPT_INJECTION]: The skill architecture creates a surface for indirect prompt injection by ingesting untrusted data into the AI agent's processing context.
Ingestion points: Untrusted external data (e.g., push notification payloads, location updates) enters the system context through the input and error fields in templates/src/services/background-tasks/background-job.ts.
Boundary markers: No explicit delimiters or instructions to ignore embedded commands are included in the default templates.
Capability inventory: Background jobs are designed to perform sensitive operations such as network requests and database updates.
Sanitization: While the documentation recommends using third-party validation libraries, the skill's base templates do not provide built-in sanitization for external inputs.

background-tasks