41 lines
1.4 KiB
Markdown
41 lines
1.4 KiB
Markdown
|
# Sprint 132 Plan: MCP Runtime Fingerprinting and Binary Freshness Guarantees
|
||
|
|
|
||
|
|
## Context
|
||
|
|
|
||
|
|
Even with correct source wiring, stale binaries can expose older MCP surfaces.
|
||
|
|
Sprint 132 introduces runtime fingerprints to guarantee clients know what binary
|
||
|
|
capabilities they are talking to.
|
||
|
|
|
||
|
|
---
|
||
|
|
|
||
|
|
## Goals
|
||
|
|
|
||
|
|
1. Fingerprint MCP binaries with deterministic capability hashes
|
||
|
|
2. Detect stale runtime vs current workspace source at startup
|
||
|
|
3. Provide explicit freshness checks for automation and agents
|
||
|
|
4. Prevent silent downgrade to older tool surfaces
|
||
|
|
|
||
|
|
---
|
||
|
|
|
||
|
|
## Steps
|
||
|
|
|
||
|
|
### Step 1559: MCP runtime fingerprint schema (10 tests)
|
||
|
|
### Step 1560: Capability hash generator from manifest + build metadata (10 tests)
|
||
|
|
### Step 1561: Workspace-source freshness comparator model (10 tests)
|
||
|
|
### Step 1562: Startup stale-binary detector integration (10 tests)
|
||
|
|
### Step 1563: `whetstone_get_runtime_fingerprint` MCP tool (8 tests)
|
||
|
|
### Step 1564: `whetstone_check_runtime_freshness` MCP tool (8 tests)
|
||
|
|
### Step 1565: Runtime freshness warning/escalation policy (8 tests)
|
||
|
|
### Step 1566: Binary capability packet export model (8 tests)
|
||
|
|
### Step 1567: Runtime freshness report artifact (8 tests)
|
||
|
|
### Step 1568: Sprint 132 integration summary + regression (8 tests)
|
||
|
|
|
||
|
|
---
|
||
|
|
|
||
|
|
## Architecture Gate
|
||
|
|
|
||
|
|
- Fingerprints must be deterministic and reproducible
|
||
|
|
- Freshness checks must include explicit source and binary provenance
|
||
|
|
- Warnings/escalations must be machine-readable
|
||
|
|
- Max 600 lines per header
|