feat(skills): add hindsight-memory skill for persistent agent memory

[benfrank241]

Summary

Proposing a new agent skill that teaches sandboxed agents how to use Hindsight for persistent memory across ephemeral sandbox sessions.

Sandboxes are isolated and disposable — when a sandbox is destroyed, everything the agent learned is lost. Hindsight solves this by providing a structured memory API that agents can call from inside the sandbox to recall past context and store new learnings.

What the Skill Covers

• Recall before work — query past context before starting tasks
• Retain after work — store learnings immediately (sandboxes can be destroyed anytime)
• Reflect for synthesis — get synthesized answers from multiple memories
• Bulk file ingestion — retain logs, reports, and investigation notes
• Cross-sandbox continuity — pick up where previous sessions left off

Proposed Files

File	Purpose
.agents/skills/hindsight-memory/SKILL.md	Main skill documentation with setup and workflows
.agents/skills/hindsight-memory/cli-reference.md	Hindsight CLI command reference
.agents/skills/hindsight-memory/example-policy.yaml	Network policy template for Hindsight API access

Context

• The NemoClaw version adapts the skill to NemoClaw's environment: nemoclaw CLI commands, NemoClaw blueprint/onboard flow, and the explicit rules policy format used in openclaw-sandbox.yaml
• Hindsight CLI commands inside the sandbox are identical to the OpenShell version

Happy to adjust scope or format based on maintainer feedback.