The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill facilitates the discovery and installation of third-party agent skills through the npx skills add <owner>/<repo> -y command. This mechanism allows the agent to fetch and execute code from arbitrary GitHub repositories. Although the instructions specify that these must be presented to the user for selection via a multi-select panel rather than being auto-installed, it enables a path for executing remote code from unverified sources discovered during the codebase scan.
[INDIRECT_PROMPT_INJECTION]: The skill's primary function involves scanning and processing the entire project codebase, including source files, configuration manifests, and CI scripts. This represents a significant ingestion point for untrusted data. Maliciously crafted instructions embedded within the code could attempt to influence the agent's behavior or output during the audit process.
Ingestion points: Full codebase audit including source code, manifests, and CI configuration files.
Boundary markers: The skill uses structured templates for output but does not define explicit boundary markers or 'ignore' instructions for the processed codebase content.
Capability inventory: The skill has access to shell execution via Bash, file writing via Write, file editing via Edit, and skill installation via npx.
Sanitization: There is no evidence of content sanitization or filtering for the data ingested from the audited files.
[EXTERNAL_DOWNLOADS]: The skill performs web searches and fetches data from external registries to identify relevant tools and skills for the project's technology stack. This is used to build a discovery set and provide recommendations for external integrations.
[COMMAND_EXECUTION]: The skill utilizes shell commands such as git log --oneline and npx skills to analyze project history and manage project-level extensions. It relies on the Bash tool to perform these system-level operations.

audit