The Agent Skills Directory

[PRIVILEGE_ESCALATION]: The skill provides installation instructions in references/code-quality-tools.md that use the sudo command to move a downloaded binary to a system directory (/usr/local/bin/), which is a high-risk privilege escalation pattern.
[EXTERNAL_DOWNLOADS]: The skill recommends downloading and installing binaries from non-trusted external GitHub repositories, such as Wilfred/difftastic and evilmartians/lefthook, which are not verified vendors. The instructions involve downloading compressed archives and executing the contents.
[METADATA_POISONING]: The skill exhibits inconsistent naming and branding; while the author is ninjasln-labs, the documentation frequently references Netresearch standards and links to a repository under the netdelegated-research organization. This inconsistency may mislead users regarding the skill's origin and safety.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data, including pull request discussion threads, review comments, and git commit history, using the gh CLI and GraphQL API. It lacks boundary markers or sanitization for this ingested data. Given the skill's capabilities to perform significant repository actions like merging, rebasing, and pushing code, this creates a risk of the agent executing malicious instructions embedded in external content.
[COMMAND_EXECUTION]: The skill utilizes powerful git features such as bisect run and rebase --exec, which allow for the execution of arbitrary shell scripts or commands. These capabilities provide a significant attack surface if the agent is influenced by malicious input.

git-workflow