Retire stale .github/WORKING_WITH_AGENTS.md (superseded by learn/agentos)

Context

.github/WORKING_WITH_AGENTS.md is an early-era relic (last substantive edits Sep–Nov 2025), written for the single-agent / Gemini-CLI / no-auto-mode days. It now contradicts its own sibling .github/AGENT_ARCHITECTURE.md (which already moved to "Autonomous Application Engine / Asynchronous Agency / Swarm") and is superseded by the canonical agent docs in learn/agentos/ (StrategicWorkflows.md, SwarmIntelligence.md, MemoryCore.md, …) plus AGENTS.md.

Live latest-open sweep: checked the latest 25 open issues on 2026-06-08; no equivalent ticket.

The Problem

The doc encodes obsolete framing: agents as "fallible junior developers requiring oversight" (now equal-peer maintainers), "never use auto-approval" (auto mode now exists, gated by the critical-gates + human-merge-gate), the human manually verifying every add_memory (now mechanized via the persist hook), the "childhood→adulthood→dementia" session-lifetime (now Agent OS memory continuity: Memory Core, session-sunset, context-recovery), and "fully automated workflows: not reliably… human-in-the-loop is the only way" (now the cross-family swarm + A2A + gated-RSI trajectory). It predates the named cross-family maintainer swarm and the Agent OS entirely.

The Fix

• Delete .github/WORKING_WITH_AGENTS.md.
• Redirect its 3 live inbound references → learn/agentos/StrategicWorkflows.md (the human-facing workflows guide): CONTRIBUTING.md, .github/AI_QUICK_START.md (×2 — also fixes the already-broken empty () link at L336), and learn/blog/context-engineering-done-right.md.
• The durable safety wisdom (human is final authority on merges/credentials; verify before trusting; deny panic/destructive actions) already lives in AGENTS.md + the critical-gates + StrategicWorkflows.md, so no content is lost.

Acceptance Criteria

• .github/WORKING_WITH_AGENTS.md deleted.
• No live inbound references remain (git grep WORKING_WITH_AGENTS returns only archived/synced content).
• Redirects point to learn/agentos/StrategicWorkflows.md; the formerly-broken AI_QUICK_START.md link is fixed.

Out of Scope

• .github/AI_QUICK_START.md is explicitly NOT retired. It is current + multi-harness (Gemini CLI / Antigravity / Claude Desktop+Code, A2A mailbox, .neo-ai-data convention) and load-bearing — README.md, the KB MCP server's openapi.yaml healthcheck guidance, the neo-identity-update skill's identity-surface ledgers, and ADR 0018 all reference it — with no learn/ duplicate. If .github-doc consolidation becomes a stated goal, RELOCATING it to learn/ is a separate careful PR (updating those load-bearing refs + a tree.json nav entry), not a retire.

Related

• Refs #12696 (v13 release epic)

Origin Session ID: 31325e4f-0fdf-4627-96ed-7983dd1b1002