The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it processes untrusted external data that could influence the agent's output logic.
Ingestion points: The dependency discovery logic in resources/implementation-playbook.md reads project manifest files such as package.json, requirements.txt, and Gemfile from the local filesystem.
Boundary markers: There are no explicit delimiters or boundary markers used in SKILL.md or the playbook's PR generation logic to isolate untrusted package data from the agent's instructions.
Capability inventory: The skill has the capability to perform network requests using requests and fetch, and execute system commands through the provided Bash remediation scripts.
Sanitization: The code does not perform sanitization or escaping of package names, versions, or descriptions before interpolating them into markdown templates for reports and pull request bodies.
[EXTERNAL_DOWNLOADS]: The skill fetches data from well-known technology services and official package registries for auditing purposes.
Evidence: Python and JavaScript snippets in resources/implementation-playbook.md communicate with registry.npmjs.org, pypi.org, rubygems.org, ossindex.sonatype.org, and bundlephobia.com to retrieve vulnerability and package metadata.
[COMMAND_EXECUTION]: The skill includes code templates designed to execute package management and testing commands on the host system.
Evidence: The playbook contains a Bash script that utilizes npm audit fix, npm update, and pip install to apply security patches and updates.

dependency-management-deps-audit