The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to analyze untrusted content from a repository (READMEs, code, scripts, and documentation) to generate verification instructions and shell commands.
Ingestion points: SKILL.md instructions require the agent to "Interview the repo" and identify user-facing features from "routes, commands, menus, or docs".
Boundary markers: The instructions lack specific delimiters or guidance to ignore potentially malicious instructions embedded within the analyzed repository files.
Capability inventory: The agent is empowered to execute repository-defined shell commands (such as package scripts or Makefile), modify the codebase to "fix" builds, and execute generated instructions.
Sanitization: The skill does not prescribe any sanitization or validation of the data extracted from the repository before it is used to generate or execute commands.
[DYNAMIC_EXECUTION]: The skill requires the agent to execute code blocks and instructions that it has dynamically generated during the same session.
Evidence: SKILL.md Section 4 ("Prove the generated skill") instructs the agent to "Run the feature file's recipe block as written, extracted from the Markdown" to verify the generated skill before delivery.
[COMMAND_EXECUTION]: The skill performs system environment checks and executes project-specific build and test scripts.
Evidence: SKILL.md Section 1 instructs the agent to run the repository's own dev commands (e.g., package scripts, Makefile) and system check commands like ls /lib/modules/$(uname -r) and findmnt /var/lib to diagnose environment issues.

create-verification-skill