Skip to content

Add AI-agent operating manual, reference docs, skills, rules, and guardrails - #3850

Merged
a-TODO-rov merged 2 commits into
mainfrom
add-ai-guidelines
Jul 17, 2026
Merged

a-TODO-rov merged 2 commits into
mainfrom
add-ai-guidelines

Conversation

@a-TODO-rov

@a-TODO-rov a-TODO-rov commented Jul 17, 2026 •

Copy link
Copy Markdown
Collaborator

Makes the repository AI-agent-friendly by adding an in-repo guidance system that also
serves human contributors: an operating manual, reference docs, task skills, coding
rules, and an enforced guardrail. Documentation and tooling only — no production or
public-API changes.

Key Decisions & Assumptions

  • One tool-neutral source under .agents/ (docs, rules, skills); tool-specific
    files (CLAUDE.md, .claude/) are thin bridges/symlinks into it, so each fact is
    edited in one place.
  • Agent material stays out of the published docs site — only user-facing pages
    remain under docs/ (MkDocs); agents read the .agents/ copies.
  • Single-owner graph: each topic has one owner file; everything else links.
  • Humans own outward/irreversible actions: AGENTS.md forbids agents from
    committing/pushing or creating/commenting on PRs and issues (a written guardrail);
    the add-a-command skill stops for human review before code generation.

Behavioral / Conceptual Changes

  • AGENTS.md operating manual + CLAUDE.md that imports it (@AGENTS.md).
  • Reference docs under .agents/docs/: architecture (command flow, code generation,
    cluster broadcast/aggregation, connections/threading/errors), a code-generation
    runbook, integration-testing notes (base+overload test pattern), and Javadoc
    conventions.
  • Skills: adding-a-redis-command, writing-javadoc, draft-pr-description.
  • Coding rules under .agents/rules/ (Java style), bridged into .claude/rules/.
  • Public docs/integration-testing.md: single-test example corrected to -Dit.test.

Note

Low Risk
Documentation and agent configuration only; the sole user-facing doc tweak is a Maven flag correction with no runtime impact.

Overview
Adds an in-repo AI-agent guidance system (documentation and tooling only — no client/runtime changes): root AGENTS.md operating manual, CLAUDE.md that imports it, and a tool-neutral tree under .agents/ (reference docs, Java rules, task skills) with .claude/ as thin bridges/symlinks.

Reference docs cover command flow and codegen pitfalls (templates vs generated APIs, sync proxy vs reactive impl), a safe API-generator runbook, integration-test notes (Failsafe -Dit.test, base+overload pattern), and Javadoc house style. Skills spell out end-to-end workflows: add a Redis command (with a human review stop before generation), write Javadoc, and draft PR text to prDescription.md.

Guardrails in AGENTS.md forbid agents from commit/push and from GitHub PR/issue actions; prDescription.md is gitignored. Published contributor docs get one fix: the “run one integration test” example in docs/integration-testing.md now uses -Dit.test instead of -Dtest.

Reviewed by Cursor Bugbot for commit 0c54df7. Bugbot is set up for automated code reviews on this repo. Configure here.

@a-TODO-rov
a-TODO-rov merged commit 771b89d into main Jul 17, 2026
8 of 9 checks passed
@a-TODO-rov
a-TODO-rov deleted the add-ai-guidelines branch July 17, 2026 15:00
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant

⌂ 换网址