docs(mcp): name the accepted seed kinds on trace_application_feature (#8220)

* docs(mcp): name the accepted seed kinds on trace_application_feature

The compact input profile reduces this schema to shallow properties, which
drops the seed union and its kind enums, so agents could not see which
kinds a seed accepts without reading the source. Describe the seed field
with its kinds, match defaults and one example; shallow reduction keeps
per-property descriptions, so the guidance survives the compact profile.

Closes #8205

* test(mcp): isolate seed guidance catalog configuration

---------

Co-authored-by: morluto <76467478+morluto@users.noreply.github.com>
This commit is contained in:
Chiragandmorluto authored and GitHub committed 2026-11-09 21:01:52 +08:00
1 parent 9108d9a627
commit 0a15246d32
2 files changed
+71 -1

No files matched your search

@@ -105,7 +105,9 @@ export const traceApplicationFeatureInputSchema = z
.strictObject({
application: evidenceSchema,
native_observations: z.array(evidenceSchema).default([]),
seed: applicationFeatureSeedSchema,
seed: applicationFeatureSeedSchema.describe(
'Typed literal starting point. kind is one of node-id, route, string, channel, module, native-export, or api; value is the literal text. match defaults to contains for string seeds and exact otherwise; an empty value requires match exact. api seeds may add entry_path. Example: {"kind":"module","value":"search.js","match":"exact"}.',
),
semantic_page: z
.strictObject({
offset: z.int().nonnegative().default(0),