41 lines
1.5 KiB
Markdown
41 lines
1.5 KiB
Markdown
|
# Sprint 131 Plan: MCP Tool Surface Source-of-Truth and Manifest Sync
|
||
|
|
|
||
|
|
## Context
|
||
|
|
|
||
|
|
Recent sprints added many MCP tools, but tool visibility can diverge across source,
|
||
|
|
build outputs, and runtime binaries. Sprint 131 makes tool-surface definition
|
||
|
|
single-source and verifiable.
|
||
|
|
|
||
|
|
---
|
||
|
|
|
||
|
|
## Goals
|
||
|
|
|
||
|
|
1. Define canonical tool manifest generation from MCP server source
|
||
|
|
2. Eliminate drift between registered tools and runtime tool listings
|
||
|
|
3. Add deterministic checks that fail builds on surface mismatch
|
||
|
|
4. Expose machine-readable tool surface metadata for downstream pipelines
|
||
|
|
|
||
|
|
---
|
||
|
|
|
||
|
|
## Steps
|
||
|
|
|
||
|
|
### Step 1549: Canonical MCP tool manifest model (10 tests)
|
||
|
|
### Step 1550: Static registry scanner for `register*Tools` blocks (10 tests)
|
||
|
|
### Step 1551: Runtime `tools/list` snapshot normalizer (10 tests)
|
||
|
|
### Step 1552: Source-vs-runtime drift comparator model (10 tests)
|
||
|
|
### Step 1553: `whetstone_export_tool_manifest` MCP tool (8 tests)
|
||
|
|
### Step 1554: `whetstone_compare_tool_surfaces` MCP tool (8 tests)
|
||
|
|
### Step 1555: Tool manifest build-gate integration (8 tests)
|
||
|
|
### Step 1556: Tool-surface checksum packet model (8 tests)
|
||
|
|
### Step 1557: Tool-surface diagnostics report artifact (8 tests)
|
||
|
|
### Step 1558: Sprint 131 integration summary + regression (8 tests)
|
||
|
|
|
||
|
|
---
|
||
|
|
|
||
|
|
## Architecture Gate
|
||
|
|
|
||
|
|
- Canonical manifest must be deterministic for identical source trees
|
||
|
|
- Source/runtime comparisons must be order-insensitive but stable
|
||
|
|
- Drift output must be structured, not free-text-only
|
||
|
|
- Max 600 lines per header
|