Dogfood: add thin caller + AGENTS.md so agents work on the agents repo itself

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

AGENTS.md

@@ -0,0 +1,36 @@
+# AGENTS.md — how changes are made in this repo
+
+Read by AI agents (via opencode) **and** humans. Follow it.
+
+## What this repo is
+The **shared AI-agent workflow** (`.gitea/workflows/agent.yml`, reusable via `workflow_call`) used by
+every other repo through a thin caller. **Changes here affect ALL repos that call it** — so be
+extra careful, keep changes small and independently mergeable, and never break the trigger gate or
+the loop guards.
+
+## Golden rules
+- You may edit **any file**. **NEVER push to `main`, NEVER merge a PR** — all work goes on a branch
+  and becomes a PR a human reviews and merges.
+- **Never print, exfiltrate, or invent secret values.**
+- Keep changes **minimal** and match the conventions already in the file you're editing.
+- Do the work on a **branch** — never paste code or diffs into the issue thread.
+
+## Branches & pull requests
+Start on `ai/issue-<N>`. Split independent changes into separate branches (one PR each). Commit and
+push incrementally. Do NOT open PRs yourself (automated). End your reply with the PR description
+block:
+
+```
+BEGIN_PR_DESCRIPTION
+## Summary
+<what changed and why>
+## Changes
+<short bullet list>
+END_PR_DESCRIPTION
+```
+
+## Understanding the "why"
+`git log` / `git show <commit>` / `git blame` — history is the changelog.
+
+## When unsure
+Make **no** changes and reply with specific questions instead.