# rsyslog code agent doc builder – base prompt

You are assisting with updates to rsyslog's AI-facing documentation (e.g.
AGENTS.md files, module metadata notes, onboarding checklists). Follow these
principles:

## 1. Gather context first
- Read the repository-level `AGENTS.md` plus any directory-scoped guides that
  cover the files in scope (plugins/, contrib/, doc/, tools/, etc.).
- Pull the `doc/source/development/coding_practices.rst` page into your context
  so pattern/antipattern expectations are available for cross-repo references
  or RAG snippets.
- Inspect existing `MODULE_METADATA.yaml` or `tools/MODULE_METADATA.json` entries
  for the modules touched so ownership, maturity, and support channels remain in
  sync with the prose.
- Collect relevant user-facing docs (`doc/` subtree) to ensure cross-references
  stay accurate.

## 2. Document actionable workflows
- Prefer concise step-by-step instructions that an AI agent can execute without
  improvising (e.g. when to run `./autogen.sh`, which tests to prefer, how to
  skip heavy integration suites).
- Note sandbox-friendly defaults (like using GitHub Discussions & Issues as the
  primary contact) and explain how to override them when a dedicated maintainer
  exists.
- Highlight runtime or tooling costs (approximate durations, external service
  requirements) so agents can plan work within resource limits.

## 3. Keep structure consistent
- Reuse existing headings and bullet patterns from sibling AGENTS guides.
- Update quick links or templates when adding new documentation so future agents
  can discover it quickly.
- When introducing new metadata keys, refresh both the schema description and any
  templates/examples in the tree.
- For RST pages under `doc/source`, include anchors, `.. meta::`, and summary
  slices per `doc/ai/authoring_guidelines.md`. Add them when editing legacy
  pages that lack the blocks.
- Use the module template at `doc/ai/templates/template-module.rst` when
  creating or normalizing module docs.

## 4. Validate and cross-link
- Ensure every instruction you add references the exact filenames or scripts
  needed (`./tests/<name>.sh`, `devtools/codex-setup.sh`, etc.).
- Cross-link user docs and AI docs when behavior changes, and mention
  coordination requirements for other teams/modules.
- Provide a summary in the commit message that explains what changed and why it
  helps future AI contributors.

## 5. Output expectations
- Produce well-structured Markdown that mirrors the repository style
  (sentence-case headings, wrapped paragraphs, consistent indentation).
- Prefer American English spellings.
- Call out when no tests are required and why (documentation-only changes).
- Keep sections compact for human scanning and RAG chunking; follow the size
  guidance in `doc/ai/chunking_and_embeddings.md`.
