Sierra Catalina proposes Context Layer: a six-step user-controlled protocol for sharing minimal context across AI agents
sierracatalina · x · 2026-09-11
Sierra Catalina published a public working proposal for a "Context Layer" — a user-controlled contract for moving the minimum useful context across models, agents, and applications.
The protocol defines one boundary and six recorded steps, with every exchange starting from a declared purpose and ending with a receipt:
- Capture: record source event, time, identity, and payload digest before interpretation
- Normalize: derive structured claims while retaining provenance, timestamps, and confidence
- Vault: encrypted originals and policy state stay inside user-controlled authority
- Decide: allow, reduce, approve, or deny based on purpose, recipient, selectors, actions, and expiry
- Bundle: issue recipient-bound, expiring, single-use context with restrictions
- Act: verify capabilities, execute, and append a minimized receipt
The published proof includes validated v0.2 contracts, an encrypted local vault, four-state policy, scoped bundles, a files adapter, and a local-agent consumer — explicitly not claiming production key custody. An interactive demo runs a synthetic request through policy, disclosure, and receipt.
More from coding & agent
- ZeroChan MCP lets AI assistants browse anime image board via Zerochan API — modelcontextprotocol · 2026-09-11
- Alpha Vantage ships MCP server for stock market data and fundamentals — modelcontextprotocol · 2026-09-11
- Codex tips: disable history and memories, start at Astra Low and raise effort — alexcovo_eth · 2026-09-11
- Indie dev seeks a tiered alerting stack for agent notifications: SMS for P1, calls for P0 — yihui_indie · 2026-09-11
- DeepsecBench security leaderboard: GPT-6 Astra tops at 37.79, Opus 5 costs $128 per run — JohnPhamous · 2026-09-11
- One Dollar Audit Offers AI Smart Contract Security Audits for $1 on Base — seanwbren · 2026-09-11