The Agent Skills Directory

[COMMAND_EXECUTION]: The skill generates shell commands that incorporate user-provided prompts (e.g., omp -p "<prompt>"). The use of double quotes for the placeholder is insufficient to prevent shell command substitution; an attacker could provide a prompt containing $(...) or backticks to execute arbitrary commands on the host machine.
[INDIRECT_PROMPT_INJECTION]: The skill acts as an ingestion point for untrusted instructions that are passed to a secondary agent environment.
Ingestion points: Multiple shell command examples in SKILL.md ingest untrusted text via the <prompt>, <prompt A>, and <prompt B> placeholders.
Boundary markers: No boundary markers or 'ignore' instructions are used to separate the delegated prompt from the execution environment.
Capability inventory: The skill utilizes shell command execution for the omp CLI, jq for JSON processing, and python3 for data extraction.
Sanitization: There is no evidence of input validation, escaping, or sanitization before interpolation into shell commands.
[DYNAMIC_EXECUTION]: The skill uses python3 -c to execute a short Python script for parsing session IDs from JSON output. While the script itself is a fixed template, it represents the execution of logic passed via command-line arguments.

omp-delegate