The Agent Skills Directory

[COMMAND_EXECUTION]: The skill extensively uses command-line tools including the Helmor binary (helmor), Git, and GitHub CLI (gh). It provides specific workflows for managing branches (git branch), resetting workspace state (git reset --hard), and staging files (git checkout, git rm). These commands are integral to the skill's primary purpose of code and workspace management.
[DYNAMIC_EXECUTION]: The skill includes a local Python script scripts/render_stack.py and instructs the agent to execute it to render JSON stack specifications into ASCII diagrams. The script is a self-contained utility that processes data from standard input or local files without using dangerous execution functions like eval() or performing network operations.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data during its 'Analyze the diff' and 'Investigate prior art' steps, specifically reading outputs from git diff, git name-status, and gh pr list. While this presents a potential surface for indirect prompt injection via malicious file names or commit messages, the risk is mitigated by the highly structured nature of the developer workflows and the lack of automated, unreviewed execution of instructions found within that data.

helmor-cli