The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes bash commands parsed from the 'Workflow' sections of external files matching the verify-*.md pattern. This behavior is central to its purpose but facilitates the execution of arbitrary shell commands defined in those local files.
[DYNAMIC_EXECUTION]: The skill dynamically parses instruction sets from local Markdown files at runtime and executes them using system tools (Grep, Glob, Read, Bash) or by spawning subagents using the Task tool. It uses string parsing to extract and run logic defined in external files.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a vulnerability surface for indirect prompt injection because it ingests content from multiple local files (.claude/skills/verify-*/SKILL.md) and interpolates their workflows into its own execution context or into subagent prompts.
Ingestion points: Relative paths defined by the naming pattern .claude/skills/verify-*/SKILL.md.
Boundary markers: The skill does not implement explicit boundary markers or 'ignore embedded instructions' headers when processing the content of the target verify skills.
Capability inventory: The skill has the capability to execute bash commands, read/write local files (during the fix phase), and spawn new agent instances.
Sanitization: There is no evidence of sanitization or validation of the commands or instructions parsed from the external verify skills before execution.

verify-implementation