The Agent Skills Directory

[External Package Execution]: The skill utilizes npx --yes deepsec@latest init to set up the environment. This downloads and runs the latest version of the DeepSec utility from the npm registry. Using official package managers for tool initialization is a standard practice in development workflows.
[Dependency Management]: It performs package installations using corepack pnpm install within the .deepsec/ directory. This ensures the environment has the necessary libraries, such as @anthropic-ai/claude-agent-sdk, to function correctly.
[Sensitive Data Handling]: The instructions include specific guidance to avoid writing credentials to configuration files and recommend using environment variables. It also includes a validation step using git diff --stat to ensure that secrets or raw scan data are not accidentally staged for version control.
[Project Data Ingestion]: To provide context for the scan, the skill reads project metadata files (like README.md and AGENTS.md) and skims source code. While this involves processing external data, it is a core requirement for the tool's primary purpose of security analysis.
Ingestion points: Project root files such as README.md, AGENTS.md, and CLAUDE.md, as well as representative source code like middleware and handlers in the /workspace/repo directory.
Boundary markers: No explicit delimiters or boundary markers are defined for the ingested file content within the instructions.
Capability inventory: The skill possesses capabilities for shell command execution (pnpm, npx, deepsec), file system read/write access within the workspace, and network access for package management.
Sanitization: The skill relies on the DeepSec tool's internal processing for handling the ingested project content; no manual sanitization steps are defined in the runbook.

deepsec