The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill fetches the official CodeQL CLI binaries from GitHub's release infrastructure (github.com/github/codeql-cli-binaries). As this is a well-known and official source for the tool, this behavior is expected and safe.
[COMMAND_EXECUTION]: The instructions involve executing shell commands to install the CLI and run the analysis. This includes wrapping the project's own build process (e.g., ./gradlew build, make all) to allow CodeQL to observe the compilation and build the semantic database.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process a codebase (--source-root=.). While this constitutes an attack surface where untrusted data (source code) enters the agent's context, the risk is inherent to the purpose of a security scanner and is mitigated by the static analysis nature of the tool.
Ingestion points: Source code files located in the project root specified in SKILL.md.
Boundary markers: None specified for the analysis process.
Capability inventory: File system access, network downloads via curl, and execution of build commands and the codeql binary.
Sanitization: Standard CodeQL database creation and analysis logic.

codeql-queries