The Agent Skills Directory

[COMMAND_EXECUTION]: Provides various shell commands for reconnaissance (e.g., npm view, pip install dry-run, gem search) and connectivity testing (e.g., curl, node, python) to verify the existence or claimability of internal package names.
[EXTERNAL_DOWNLOADS]: References external security research tools and repositories on GitHub, including tools from well-known security entities such as Visma and Synacktiv.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to analyze project manifests, which constitute untrusted external data. 1. Ingestion points: Manifest files like package.json, requirements.txt, and Gemfile. 2. Boundary markers: Explicit disclaimer requiring authorized access and testing environments. 3. Capability inventory: Shell reconnaissance commands and network callback proof-of-concept snippets. 4. Sanitization: Not applicable as the skill provides methodology for guided review rather than automated data processing.

dependency-confusion