The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes untrusted GitHub Action workflow files which could contain malicious instructions designed to mislead the agent during the validation process (Indirect Prompt Injection).
Ingestion points: Reads content from .github/workflows/*.yml for validation.
Boundary markers: No delimiters or ignore instructions are used when reading or processing the workflow files.
Capability inventory: The skill executes shell commands via actionlint and reports results back to the agent context.
Sanitization: There is no sanitization or escaping of the workflow content before it is processed or reported.
[EXTERNAL_DOWNLOADS]: The skill downloads the actionlint binary from GitHub's official repository via the releases API (api.github.com/repos/rhysd/actionlint/releases/latest). This is a well-known service and the source is appropriate for the skill's purpose.
[COMMAND_EXECUTION]: The skill performs system-level modifications by attempting to install the actionlint binary into /usr/local/bin using tar and curl commands. While standard for tool installation, this modifies the system environment.

actionlint