The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill instructs the user to download a shell script from a remote URL and pipe it directly into a shell for execution: curl https://raw.githubusercontent.com/nektos/act/master/install.sh | sudo bash.
[PRIVILEGE_ESCALATION]: The remote script execution for the 'act' tool is performed using sudo, granting the unverified remote script full administrative access to the system.
[EXTERNAL_DOWNLOADS]: The skill pulls code and installation scripts from external sources, including githubusercontent.com (nektos/act) and github.com (rhysd/actionlint via go install), which are not within the verified trust scope.
[INDIRECT_PROMPT_INJECTION]: The skill establishes a vulnerability surface by processing repository workflow files using external tools without explicit sanitization or boundary markers.
Ingestion points: Reads files matching .github/workflows/*.yml (SKILL.md).
Boundary markers: None identified; tools are run directly on local files.
Capability inventory: Executes gh (GitHub CLI), act (Docker-based workflow execution), and various linters (SKILL.md).
Sanitization: None mentioned for the data processed by actionlint or yamllint.

github-actions