The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute several shell commands for testing, linting, and synchronizing documentation within the repository environment.
Evidence: Commands include pnpm run test:docs, pnpm run doc-sync, pnpm run lint, pnpm run verify-translation-pairing, pnpm run verify-doc-budgets, pnpm run verify-md-links, pnpm docs:dev, pnpm docs:check, dsh plugin --profile <name> add <package>, git fetch origin, and git diff --check.
[INDIRECT_PROMPT_INJECTION]: The skill functions by ingesting and acting upon content from repository files, which constitutes a surface for indirect prompt injection if those files contain malicious instructions.
Ingestion points: Found in SKILL.md (Workflow Step 1) and references/review.md (Evidence review), where the agent is directed to read target documentation pages, source code (src/index.ts), and package manifests (package.json).
Boundary markers: The skill utilizes YAML frontmatter and specific Markdown templates (e.g., templates/package-reference.md) to structure documentation, which provides some context separation but does not prevent instruction injection within those fields.
Capability inventory: The skill has the capability to run significant shell commands (pnpm, dsh) and perform git operations, which could be abused if the agent follows instructions embedded in the documents it audits.
Sanitization: There are no explicit mechanisms mentioned for sanitizing or escaping the content read from Markdown or code files before it is processed by the agent.

dsh-doc