The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill downloads and executes the dsh-doctor CLI from the npm registry using npm exec or npm install. These resources are provided by the skill author (@bruc3van).\n
Evidence: npm view @bruc3van/dsh-doctor version, npm exec --package=@bruc3van/dsh-doctor@<version>, and npm install --global @bruc3van/dsh-doctor in references/cli-bootstrap.md.\n- [COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands for Git operations, environment inspection, and DSH plugin management.\n
Evidence: Commands such as git diff, npm view, and dsh-doctor migrate analyze are used throughout SKILL.md and references/source-investigation.md.\n- [DYNAMIC_EXECUTION]: During the verification phase, the skill executes project-defined lifecycle scripts to confirm the success of the migration.\n
Evidence: The agent runs typecheck, build, test, and pack:check scripts as defined in the plugin's package.json (documented in references/verification.md).\n- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted plugin source code and configuration files, which represents a potential attack surface for indirect prompt injection.\n
Ingestion points: The plugin repository root, including source files (.ts, .js) and metadata files (package.json, manifests) as specified in SKILL.md.\n
Boundary markers: The skill includes specific logical gates for developer authorization but does not use explicit prompt delimiters to isolate content from the plugin files.\n
Capability inventory: The skill has the capability to write to the file system (migrate apply), execute shell commands (npm, git), and download packages via the npm registry.\n
Sanitization: No explicit sanitization or filtering of plugin content is mentioned; the skill relies on the internal logic of the dsh-doctor CLI and developer oversight of the migration plan.

dsh-plugin-upgrade