laravel-cloud-ops
Warn
Audited by Gen Agent Trust Hub on Oct 11, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONDYNAMIC_EXECUTIONDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [DYNAMIC_EXECUTION]: The skill explicitly instructs the agent to execute larger scripts by base64-encoding them locally, writing the decoded content to a temporary file on the remote server, and then executing it using PHP.
- Evidence: The pattern
echo <b64> | base64 -d > /tmp/x.php && php /tmp/x.phpis suggested for script execution inSKILL.md. - [COMMAND_EXECUTION]: The skill facilitates the execution of arbitrary commands on remote Cloud environments and relies on a local binary in a non-standard directory for API interactions.
- Evidence: Use of
cloud command:run <env> --cmd="..."andcloud tinkerfor remote execution. - Evidence: The
cloudapibash helper depends on the local script$HOME/.dotfiles/bin/spatie-ops. - [DATA_EXFILTRATION]: The skill provides instructions for accessing sensitive information such as environment variables and bucket secret keys. While it includes warnings against printing secrets in the conversation, the presence of commands like
cloud env:get --show-sensitivecreates a risk of accidental or coerced credential exposure. - Evidence: Accessing
access_key_secretvia/buckets/{id}/keysand retrieving environment variables viacloud env:get <env> --json. - [INDIRECT_PROMPT_INJECTION]: The skill ingests data from external sources (Laravel Cloud API and CLI outputs) and possesses high-privilege capabilities (remote command execution, environment modification), making it a candidate for indirect injection attacks if outputs are not properly sanitized.
- Ingestion points: Responses from
cloudapi, logs retrieved viacloud env:logs, and output fromcloud command:runinSKILL.md. - Boundary markers: The skill references an external
spatie-opsskill for "safe API output" but does not specify internal validation logic. - Capability inventory: The skill can execute remote commands, run interactive tinker sessions, modify environment variables, and manage database clusters.
- Sanitization: Instructions suggest masking env vars and redacting sensitive values from API responses, though this relies on the correct behavior of the external
spatie-opstool.
Audit Metadata