The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill exposes an attack surface through the $ARGUMENTS placeholder and the ingestion of external project data.
Ingestion points: SKILL.md contains an $ARGUMENTS placeholder in the 'Imported: Requirements' section. Additionally, the skill's primary function is to process untrusted external data from dependency manifests such as package.json, requirements.txt, and lock files.
Boundary markers: The skill lacks explicit delimiters or instructions to ignore embedded commands within the interpolated arguments or the project files it analyzes.
Capability inventory: The skill (via implementation-playbook.md) includes capabilities for file system access, network communication (via Python requests), and the generation/execution of shell scripts for package management.
Sanitization: No sanitization or validation logic is present to filter malicious instructions from the ingested content.
[COMMAND_EXECUTION]: The resources/implementation-playbook.md file provides operational templates that instruct the agent to execute shell commands. Examples include npm audit fix, npm update, pip install, and pip-compile for automated remediation of vulnerabilities.
[EXTERNAL_DOWNLOADS]: The skill contains Python and JavaScript code snippets that perform network operations to fetch metadata from external services.
Registry Queries: Fetches vulnerability data from registry.npmjs.org, pypi.org, rubygems.org, and ossindex.sonatype.org.
Performance Data: Fetches package size statistics from bundlephobia.com.

dependency-management-deps-audit