The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill instructs the agent to perform a "cold start" to learn how to build and run a project, and then "persist what you learned" by creating a new instruction file at .claude/skills/verify/SKILL.md. It specifically directs the agent to "invoke it with the Skill tool" in subsequent sessions. This behavior constitutes dynamic generation and execution of agent instructions based on runtime findings.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from the repository environment that could contain malicious instructions designed to influence the agent's behavior.
Ingestion points: The agent reads the output of git diff, git log, and gh pr diff, as well as file contents within the repository.
Boundary markers: Absent. The instructions do not define clear delimiters or include warnings to ignore instructions embedded within the diff or pull request content.
Capability inventory: The skill uses shell commands (git, ls, tmux) and has the capability to write to the filesystem.
Sanitization: Absent. No validation or filtering is performed on the data ingested from the diffs or code files.
[COMMAND_EXECUTION]: The skill utilizes shell commands to inspect the repository and manage execution environments, including git, ls, tmux, and mktemp. These are standard tools for development and verification workflows but represent the primary capability set available to the agent.

verify