The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user data and persists it to the filesystem, creating an injection surface for other skills that might read the same data.
Ingestion points: User input regarding travel plans, destinations, and medical history used in functions 1, 3, and 6.
Boundary markers: No boundary markers or instructions to ignore embedded commands are present when writing to data/travel-health-tracker.json.
Capability inventory: The skill uses Write and Read tools to manage a local database of health records.
Sanitization: There are no instructions for the agent to sanitize or escape user-provided content before storage.
[DATA_EXFILTRATION]: The skill manages highly sensitive Personal Identifiable Information (PII) and Protected Health Information (PHI).
Evidence: The skill instructions specifically mention handling and storing "Passport numbers", "Blood Type", "Allergies", "Current Medications", and "Medical Conditions".
Context: Storing this data in shared local files (data/travel-health-tracker.json) exposes it to any other skill with Read access to the same directory.
[METADATA_POISONING]: The skill's YAML frontmatter contains an unusual and potentially misleading metadata field.
Evidence: The frontmatter includes risk: critical. This is non-standard and could be an attempt to influence how the platform or analyzer categorizes the skill's risk level outside of its actual functional behavior.

travel-health-analyzer