The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses shell_exec to run curl for authentication and secret management operations (List, Create, Patch, Delete) against a remote Infisical instance.
[PERSISTENCE]: The agent is instructed to use schedule_create to establish a recurring task named "infisical-sync", allowing the skill to maintain its activity and periodically execute synchronization cycles.
[DYNAMIC_EXECUTION]: The skill employs python3 -c to execute dynamically generated code for parsing JSON response data from the Infisical API.
[DATA_EXFILTRATION]: The skill accesses sensitive environment variables such as INFISICAL_CLIENT_SECRET and transmits vault data to an external endpoint defined by INFISICAL_URL. While this is the intended purpose, it represents a data transit path for sensitive information.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data retrieved from API responses which could contain malicious content designed to influence the agent.
Ingestion points: Secret values, keys, and workspace metadata fetched from the Infisical API in SKILL.md (Phase 4).
Boundary markers: Absent; there are no delimiters or instructions to ignore embedded commands within the fetched secret data.
Capability inventory: The agent possesses shell_exec, vault_set, file_write, and schedule_create as defined in HAND.toml.
Sanitization: Absent; secret values retrieved from the external source are interpolated directly into shell command strings (e.g., for hash computation or sync-back) without robust escaping or validation.

infisical-sync-skill