The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the agent to download and install several third-party security tools and utilities.
Fetches the Syft installation script from Anchore's official GitHub repository (https://raw.githubusercontent.com/anchore/syft/main/install.sh).
Recommends the installation of osv-scanner from Google's repository, cosign from Sigstore, and gitleaks via Go's package manager.
Uses npm to install @cyclonedx/cdxgen and snyk globally.
[REMOTE_CODE_EXECUTION]: The skill contains a pattern where an external script is downloaded and piped directly into the shell.
Evidence: curl -sSfL https://raw.githubusercontent.com/anchore/syft/main/install.sh | sh in SKILL.md for installing the Syft SBOM tool. This is a common installation pattern for this well-known security tool.
[INDIRECT_PROMPT_INJECTION]: The skill's workflow involves ingesting and processing data from external, potentially untrusted sources (vulnerability databases and automated scanner outputs).
Ingestion points: Vulnerability data from NVD, OSV.dev, and GitHub Advisory; scanner results from Trivy, Snyk, and OSV-Scanner; and SBOM files (SKILL.md, references/sbom-sca-methodology.md).
Boundary markers: Absent. The instructions do not specify the use of delimiters or warnings to ignore embedded instructions within the processed vulnerability reports or scan results.
Capability inventory: The skill uses shell commands for tool execution, performs file system reads/writes for SBOM generation, and utilizes network requests (e.g., curl to upload reports to Dependency-Track) (SKILL.md, references/cicd-pipeline-security.md).
Sanitization: Not explicitly implemented. The agent is directed to parse and analyze the content of vulnerability reports without specific sanitization instructions.

supply-chain-security