The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands including mdbook build, mdbook test, and various repository-specific validation scripts.
Evidence: Section 8 of SKILL.md defines a validation workflow that relies on executing these commands against the local filesystem.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process a wide variety of untrusted data from the repository, which could potentially contain malicious instructions intended to influence the agent's output or documentation process.
Ingestion points: The agent reads book.toml, SUMMARY.md, repository source code, tests, schemas, and architectural records.
Boundary markers: The instructions mandate the use of specific modes (Edit vs. Review) and require the agent to classify information into types (e.g., Contract, Invariant, Rationale).
Capability inventory: The skill has capabilities for file reading, file writing (in Edit mode), and shell command execution.
Sanitization: The instructions tell the agent to inspect the final diff for unrelated edits and to use isolated temporary copies for validation if side effects are suspected.
[REMOTE_CODE_EXECUTION]: By instructing the agent to run repository-provided build wrappers, preprocessors, and test suites, the skill creates a surface for executing arbitrary code defined within the repository being documented.
Evidence: SKILL.md Section 7 and 8 advise the agent to inspect "build wrappers, dependency files, and CI" and run "repository Markdown, link, spelling, style, schema, and documentation checks," which involves executing repository-local scripts.

mdbook-internals