The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill provides instructions to install the Brev CLI agent skill using a pattern that downloads a script from a remote URL and pipes it directly into the bash shell.
Evidence: curl -fsSL https://raw.githubusercontent.com/brevdev/brev-cli/main/scripts/install-agent-skill.sh | bash in SKILL.md.
[EXTERNAL_DOWNLOADS]: The skill depends on downloading external scripts and tools from the brevdev GitHub repository during setup.
[COMMAND_EXECUTION]: The skill's primary operation involves constructing and executing complex shell commands locally and on remote instances using the brev exec tool.
It specifically manages Docker lifecycles (login, pull, run, logs, rm) through shell string concatenation.
[INDIRECT_PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection due to the way it handles external data without sufficient sanitization or boundary markers.
Ingestion points: Variable interpolation of $IMG, $ARCH, $ACTION, $TIER, $RESULTS_ROOT, and $REMOTE_CMD into shell commands in SKILL.md.
Boundary markers: No delimiters or specific instructions are provided to the agent to treat these variables as untrusted data or to ignore embedded instructions.
Capability inventory: The skill utilizes the bash tool and possesses the ability to execute arbitrary commands on remote cloud instances via the Brev CLI.
Sanitization: The skill mentions a redact_secrets.py script for linting secrets, but it does not provide mechanisms for general input validation or shell escape logic for all interpolated parameters.

tao-run-on-brev