The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill contains instructions to install actionlint using 'bash <(curl ...)', a pattern that fetches and executes a script directly from a remote GitHub repository ('rhysd/actionlint'). This is a high-risk practice because it executes unverified code in the user's environment, bypassing integrity checks.
[EXTERNAL_DOWNLOADS]: The skill references several third-party GitHub Actions and tools from repositories such as 'suzuki-shunsuke/pinact-action' and 'zizmorcore/zizmor-action'. While some use pinned commit SHAs, these sources are not from the pre-approved trusted vendor list, representing an external dependency on third-party code.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection as it processes workflow files that could contain malicious instructions from untrusted contributors. Ingestion points: Reads all files under '.github/workflows/' for analysis. Boundary markers: No specific instructions or delimiters are provided to the agent to treat workflow content as untrusted data. Capability inventory: Executes shell commands and package management tasks (nix run, bash). Sanitization: There is no evidence of sanitization or content validation before the workflow data is passed to the analysis tools.

gha-lint