The Agent Skills Directory

[COMMAND_EXECUTION]: The skill relies on the shell_exec tool to run curl commands and python3 inline scripts for all API interactions and data processing tasks.
[DATA_EXFILTRATION]: The agent is instructed to access sensitive environment variables (INFISICAL_CLIENT_SECRET) and transmit them to a user-provided URL (INFISICAL_URL). It also retrieves secret values from the local vault to push them to this dynamic network endpoint.
[EXTERNAL_DOWNLOADS]: The skill is designed to download secret payloads and project metadata from a remote Infisical instance and persist them in the agent's local storage and vault.
[PROMPT_INJECTION]: The skill processes untrusted data in the form of secret keys and values fetched from a remote API, creating an indirect prompt injection surface. (1) Ingestion points: Secret values and names retrieved via curl from the Infisical API in HAND.toml (Phase 4). (2) Boundary markers: None identified for the secret content itself; the agent is simply instructed to store retrieved values. (3) Capability inventory: shell_exec, file_write, vault_set, schedule_create, and event_publish. (4) Sanitization: The skill provides instructions to redact secret values from logs but lacks explicit validation or sanitization of the secret content before it is stored or processed.

infisical-sync-skill