The Agent Skills Directory

[PROMPT_INJECTION]: The skill ingests untrusted content from multiple sources, which could be leveraged for indirect prompt injection attacks.
Ingestion points: External data enters the system through itineraryThemeText (pasted text), travelResourceFile (uploaded documents), and destinationReferenceLink (external URLs), as defined in SKILL.md and references/form-schema.json.
Boundary markers: The skill does not provide clear delimiters or instructions to the agent to treat input data as untrusted or to ignore embedded commands within these inputs.
Capability inventory: The skill's operational capability is restricted to making network requests to the ai-skills.ai API via scripts/run.py. No local command execution or file system modification capabilities were detected.
Sanitization: The skill passes parameters directly to the API in JSON format without performing input validation or sanitization for potential injection patterns.
[EXTERNAL_DOWNLOADS]: The skill's execution runner performs network operations to a remote service.
Evidence: scripts/run.py connects to https://ai-skills.ai to execute the skill's logic.
Context: This behavior is consistent with the skill's stated purpose and infrastructure.

travel-itinerary-content-planner