The Agent Skills Directory

[COMMAND_EXECUTION]: The skill's primary function is to execute an external binary (the Claude Code CLI) via shell commands.
Evidence: SKILL.md provides templates for shell execution such as printf '%s' "$TASK_PROMPT" | <launcher> --print --output-format stream-json --verbose.
[INDIRECT_PROMPT_INJECTION]: The skill provides an interface that pipes untrusted data into an agentic tool, creating a vulnerability surface for indirect prompt injection.
Ingestion points: The $TASK_PROMPT variable in SKILL.md and references/monitoring.md is used as the primary input for the automated coding agent.
Boundary markers: There are no delimiters or specific instructions to the sub-agent to ignore potential instructions embedded within the processed data.
Capability inventory: The skill explicitly enables tools like Read, Grep, and Glob in restricted mode, and potentially broader filesystem and network capabilities in default mode.
Sanitization: No sanitization, escaping, or validation logic is present for the input piped into the CLI.
[DYNAMIC_EXECUTION]: The skill dynamically assembles shell commands using variables for the launcher and the task prompt.
Evidence: The execution logic relies on string interpolation of <launcher> and $TASK_PROMPT into a shell pipeline.

claude-code-coding-agent