The Agent Skills Directory

EXTERNAL_DOWNLOADS (LOW): The installation instructions point to an untrusted GitHub repository (github.com/barneyjm/camino-skills), which is not among the verified organizations listed in the security policy.
DATA_EXFILTRATION (LOW): The skill sends user-provided locations and descriptions to an external domain (api.getcamino.ai). While this is the primary function of the skill, it constitutes a data exit point to a non-whitelisted domain.
PROMPT_INJECTION (LOW): The skill possesses a surface for indirect prompt injection as it ingests untrusted JSON data that is subsequently processed by the agent.
Ingestion points: Command-line JSON argument in scripts/travel-planner.sh.
Boundary markers: Absent.
Capability inventory: Network access via curl and JSON parsing via jq.
Sanitization: Validates JSON format via jq but does not sanitize the string content of fields like 'purpose'.
COMMAND_EXECUTION (SAFE): Shell commands in the script are properly quoted and validated, effectively preventing basic command injection vulnerabilities.
CREDENTIALS_UNSAFE (SAFE): The skill follows best practices by requiring the CAMINO_API_KEY to be set via environment variables rather than hardcoding credentials.

travel-planner