The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests and processes untrusted data from project manifest and lock files (e.g., package.json, requirements.txt, go.mod).
Ingestion points: resources/implementation-playbook.md contains logic to discover and parse various dependency manifest files from the project environment.
Boundary markers: The instructions do not specify explicit delimiters to isolate untrusted data when generating security reports or pull request descriptions.
Capability inventory: The skill can execute shell commands via remediation scripts and perform network operations to fetch metadata.
Sanitization: No explicit sanitization or escaping of manifest content (such as package names or descriptions) is shown before inclusion in agent outputs.
[EXTERNAL_DOWNLOADS]: Fetches vulnerability and package metadata from well-known public registries and services.
Evidence: Python and JavaScript code in resources/implementation-playbook.md makes network requests to registry.npmjs.org, pypi.org, rubygems.org, ossindex.sonatype.org, and bundlephobia.com.
[COMMAND_EXECUTION]: Provides automated remediation scripts that execute package manager CLI tools to address vulnerabilities.
Evidence: resources/implementation-playbook.md includes bash scripts that run npm audit fix, npm update, and pip-compile.

dependency-management-deps-audit