The Agent Skills Directory

[COMMAND_EXECUTION]: The skill defines an execution pattern in SKILL.md (Step 3) where shell commands are constructed by interpolating values from project metadata and user prompts. Specifically, the --model, --voice, and --prompt arguments are directly mapped to external inputs without escaping instructions, which could allow shell metacharacters to alter command behavior.
Evidence: node "$OD_BIN" media generate ... --model "<audioModel from metadata>" ... --prompt "<assembled prompt from Step 2>" in SKILL.md.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection (Category 8) because it processes untrusted project metadata and user-supplied scripts and uses them in a powerful subprocess execution without sanitization.
Ingestion points: Project metadata fields such as audioModel, voice, and user-provided script content used in the prompt argument are read into the agent context.
Boundary markers: There are no boundary markers or explicit instructions for the agent to ignore or escape embedded commands within the prompt data.
Capability inventory: The skill possesses the capability to execute shell commands via a subprocess call (node "$OD_BIN").
Sanitization: The instructions do not include any steps for the agent to validate, escape, or sanitize the content before it is passed to the command-line interface.

audio-jingle