The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The scripts/validate-workflows.sh script downloads a shell script from a remote GitHub repository and executes it using bash.
Evidence: curl -fsSL "https://raw.githubusercontent.com/rhysd/actionlint/v$actionlint_version/scripts/download-actionlint.bash" -o "$script" && bash "$script" "$actionlint_version" "$tmp_dir" in the resolve_actionlint function.
[EXTERNAL_DOWNLOADS]: The validation tool automatically fetches external binaries and dependencies from remote sources if they are not pre-installed.
It uses go install to fetch actionlint from github.com/rhysd/actionlint.
It uses uvx or pipx to run zizmor from external package registries.
It downloads a shell script from raw.githubusercontent.com.
[COMMAND_EXECUTION]: The skill instructs users to run a provided shell script that performs various operations on the filesystem and invokes external CLI tools.
Evidence: bash /mnt/skills/user/github-actions-ci/scripts/validate-workflows.sh [path-to-repo] in SKILL.md.
[PROMPT_INJECTION]: The validation script processes workflow files from a user-provided directory, creating an attack surface for indirect prompt injection if the files contain malicious content designed to exploit the linter or security tools.
Ingestion points: Reads all .yml and .yaml files in the specified .github/workflows directory.
Capability inventory: Invokes actionlint, zizmor, and act to analyze the contents of these files.
Boundary markers: None present.
Sanitization: None present.

github-actions-ci