The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The SKILL.md file contains instructions for a setup process that uses a piped shell command curl https://mise.run | sh. This pattern downloads a script from an external URL and executes it directly in the shell without prior inspection, which is a significant security risk if the remote source is compromised.
[PERSISTENCE]: In resources/validation-pipeline.md, the skill provides logic to automatically create or overwrite local Git hooks (commit-msg, pre-commit, pre-push). These hooks are configured to execute shell commands (mise run) every time a user performs git operations, establishing a persistence mechanism that ensures code execution outside the agent's immediate session.
[COMMAND_EXECUTION]: The script resources/affected-checks.py utilizes the subprocess.run function to execute system commands including git and mise. While it avoids using a shell environment directly, it still facilitates arbitrary command execution based on repository paths and user-provided arguments.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from the codebase, such as git diff outputs and mise.toml configuration files, to drive its control flow and determine which shell tasks to run. This creates an attack surface where malicious instructions embedded in a repository's metadata or commit history could influence the agent's command execution behavior.

oma-dev-workflow