The Agent Skills Directory

[COMMAND_EXECUTION]: The skill employs execution hooks that interpolate user-controlled variables directly into shell-like commands.
Evidence: The pre hook in SKILL.md executes memory_store "spec_start_$(date +%s)" "Task: $TASK".
Risk: If the platform evaluates these strings in a shell without proper escaping, the $TASK variable can be used to inject arbitrary shell commands.
[PROMPT_INJECTION]: The skill demonstrates a vulnerability surface for indirect prompt injection by ingesting untrusted data without boundary markers.
Ingestion points: The $TASK variable is interpolated into the pre hook in SKILL.md.
Boundary markers: There are no delimiters (e.g., XML tags or triple quotes) used to isolate the variable content.
Capability inventory: The skill uses echo and memory_store operations which, while limited, are executed automatically upon skill invocation.
Sanitization: No sanitization, escaping, or validation logic is present to filter malicious instructions or shell metacharacters from the $TASK input.

agent-specification