add report generation skills

[bgmcmullen]

Four new AI report skills added, each as a SKILL.md, registered for both the Claude and Codex client plugins:

• hypaware-ai-adoption-report — AI Adoption Profile: per-gateway utilization and multi-agent/fan-out adoption.
• hypaware-ai-improvement-report — analyzes agent logs and proposes skill / subagent / AGENTS.md|CLAUDE.md improvements.
• hypaware-ai-security-report — security & risk review of gateway logs, severity-ranked with guardrail recommendations.
• hypaware-ai-spend-report — token spend review (by user/repo/model/gateway) with waste scorecard and cost levers.

Wiring:

• Registered the four skills in both plugin manifests (claude/hypaware.plugin.json, codex/hypaware.plugin.json) and in each plugin's src/index.js activate() loop. The codex index.js was refactored from a single hypaware-query registration to a loop over all five skills.
• Updated the codex plugin description/manifest to mention the new report skills.

hypaware-query skill expanded (both clients) to support the reports:

• New "Remote queries" section documenting hyp query --remote , hyp remote add/login/list, query-scoped tokens, doubled truncation caveats, and the two ways a host's MCP can be attached (hyp --remote CLI vs. direct mcp__hypaware__* tools).
• New note on where token counts live (attributes.usage on assistant rows, with Codex/Claude differences and dedup-per-message_id guidance) — the foundation for the spend report.