The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a variety of local shell commands to analyze the codebase, including find, grep, wc, sed, jq, and npm. These commands are used to identify directory structure, count files, locate code patterns like TODOs, and check dependency counts.
[EXTERNAL_DOWNLOADS]: The skill interacts with the NPM registry (npm audit, npm outdated) and the GitHub API (gh api) to retrieve security information and configure repository settings. These interactions utilize official tools to communicate with well-known and trusted services.
[INDIRECT_PROMPT_INJECTION]: The skill processes content from the local codebase which could theoretically contain malicious instructions.
Ingestion points: The skill reads directory structures, code comments (via grep), and configuration files (via jq).
Boundary markers: None explicitly defined to isolate processed code content from the agent's instructions.
Capability inventory: The skill has the ability to read the file system, execute local shell commands, and make network requests via the GitHub CLI.
Sanitization: Output is primarily sanitized by piping results to wc -l (reducing data to counts) or using head/tail to limit the amount of text returned to the agent context.

architecture-review