The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses run_command to execute shell commands (test and rm) for file system operations.
Logic: test -f .claude/ralph-loop.local.md && echo "EXISTS" || echo "NOT_FOUND" to check file existence.
Logic: rm .claude/ralph-loop.local.md to delete the state file.
[INDIRECT_PROMPT_INJECTION]: The skill ingests data from a local file that may be influenced by other processes or tools, creating a surface for indirect instructions if that file's content is compromised.
Ingestion points: Reads .claude/ralph-loop.local.md in SKILL.md via view_file to extract the iteration: field.
Boundary markers: None identified; the iteration value is directly included in the agent's output.
Capability inventory: Uses run_command which allows arbitrary shell execution within the environment.
Sanitization: No specific sanitization or validation of the iteration value is performed before reporting it to the user.

cancel-ralph