v0.9.2 - AGENTS.md support

What's new

Adds memoryFiles config to control which memory file(s) auto-memory maintains. This enables multi-agent workflows where Claude Code, OpenAI Codex, Gemini, and other AI agents all read from the same memory file.

Three modes

memoryFiles	Behaviour
absent or ["CLAUDE.md"]	Existing behaviour unchanged (backwards compatible)
["AGENTS.md"]	AGENTS.md gets full content, CLAUDE.md untouched
["CLAUDE.md", "AGENTS.md"]	AGENTS.md gets full content, CLAUDE.md is a one-line redirect

Changes

• /auto-memory:init: New Step 1a wizard question to select memory file(s)
• /auto-memory:status: Now shows active memory file and all instances found in project
• /auto-memory:sync: Filters using memoryFiles list instead of hardcoded CLAUDE.md
• post-tool-use.py: should_track() excludes all configured memory files (not just CLAUDE.md)
• trigger.py: auto_commit_memory_files() commits whichever files are configured; spawn reason and pre-tool-use deny message reference the active file by name
• agents/memory-updater.md: Phase 4 reads config to discover active memory file (AGENTS.md or CLAUDE.md)
• skills/codebase-analyzer: Step 7 generates correct file set per config; three new templates added
• skills/memory-processor: Operates on active memory file; skips CLAUDE.md redirect files
• New templates: AGENTS.root.md.template, AGENTS.subtree.md.template, CLAUDE.redirect.md.template

Test coverage

246 tests (25 new tests covering all three memoryFiles modes, active file resolution, commit message variants, and template structure).