feat(analyst): type every agent improvement surface

CHANGELOG.md

@@ -4,6 +4,17 @@ All notable changes to `@tangle-network/agent-eval` and its sibling `agent-eval-
 
 ---
 
+## [0.112.0] — 2026-07-10 — complete agent-surface findings
+
+### Added
+
+- Added typed finding subjects for skills, MCP servers and tools, hooks, subagents, workflows, rollout policy, generic agent-profile fields, and code paths.
+- Routed every new subject into a typed policy edit so products do not need local string classifiers.
+
+### Changed
+
+- Updated the direct `@tangle-network/agent-interface` dependency from `^0.10.0` to `^0.22.0`.
+
 ## [0.111.0] — 2026-07-09 — repository-clustered paired inference
 
 ### Added

clients/python/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
 
 [project]
 name = "agent-eval-rpc"
-version = "0.111.0"
+version = "0.112.0"
 description = "Python RPC client for @tangle-network/agent-eval — judge content against rubrics over HTTP or stdio RPC. Eval logic runs in the Node runtime; this package is a thin wire client."
 readme = "README.md"
 requires-python = ">=3.10"

clients/python/src/agent_eval_rpc/__init__.py

@@ -58,7 +58,7 @@
 try:
     __version__ = version("agent-eval-rpc")
 except PackageNotFoundError:
-    __version__ = "0.111.0"
+    __version__ = "0.112.0"
 
 __all__ = [
     "Client",

package.json

@@ -1,6 +1,6 @@
 {
   "name": "@tangle-network/agent-eval",
-  "version": "0.111.0",
+  "version": "0.112.0",
   "description": "Evaluate and improve AI agents from runs, traces, judges, and feedback. Compare candidates, cluster failures, measure lift, and gate releases.",
   "homepage": "https://github.com/tangle-network/agent-eval#readme",
   "repository": {
@@ -146,7 +146,7 @@
     "@asteasolutions/zod-to-openapi": "^8.5.0",
     "@ax-llm/ax": "^19.0.25",
     "@hono/node-server": "^2.0.0",
-    "@tangle-network/agent-interface": "^0.10.0",
+    "@tangle-network/agent-interface": "^0.22.0",
     "@tangle-network/tcloud": "^0.4.14",
     "hono": "^4.12.16",
     "zod": "^4.3.6"

src/analyst/finding-subject.test.ts

@@ -1,11 +1,19 @@
 import { describe, expect, it } from 'vitest'
 import {
   FINDING_SUBJECT_KINDS,
+  FINDING_SUBJECT_SYNTAX,
   type FindingSubject,
+  findingSubjectGrammarPromptFor,
   KIND_EXPECTED_SUBJECTS,
   parseFindingSubject,
   renderFindingSubject,
 } from './finding-subject'
+import {
+  FAILURE_MODE_KIND_SPEC,
+  IMPROVEMENT_KIND_SPEC,
+  KNOWLEDGE_GAP_KIND_SPEC,
+  KNOWLEDGE_POISONING_KIND_SPEC,
+} from './kinds'
 
 describe('parseFindingSubject — knowledge loci', () => {
   it('parses agent-knowledge:wiki:<slug>', () => {
@@ -91,6 +99,30 @@ describe('parseFindingSubject — runtime surfaces', () => {
     })
   })
 
+  it.each([
+    ['skill:linear-triage', { kind: 'skill', name: 'linear-triage' }],
+    ['mcp:linear', { kind: 'mcp', server: 'linear' }],
+    ['mcp:linear:get_issue', { kind: 'mcp', server: 'linear', tool: 'get_issue' }],
+    ['hook:post-tool-use', { kind: 'hook', name: 'post-tool-use' }],
+    ['subagent:reviewer', { kind: 'subagent', name: 'reviewer' }],
+    ['workflow:linear-ticket', { kind: 'workflow', name: 'linear-ticket' }],
+    ['rollout-policy:maxTurns', { kind: 'rollout-policy', field: 'maxTurns' }],
+    ['agent-profile:model', { kind: 'agent-profile', field: 'model' }],
+    ['code:src/dispatch.ts', { kind: 'code', path: 'src/dispatch.ts' }],
+  ])('parses %s', (raw, expected) => {
+    expect(parseFindingSubject(raw)).toEqual(expected)
+  })
+
+  it.each([
+    'skill:CodeReview',
+    'mcp:Linear',
+    'hook:',
+    'subagent:',
+    'workflow:',
+  ])('rejects malformed runtime subject %s', (raw) => {
+    expect(parseFindingSubject(raw)).toBeNull()
+  })
+
   it('parses rag:<corpus>:<doc>', () => {
     expect(parseFindingSubject('rag:irs-rulings:rev-rul-2024-12')).toEqual({
       kind: 'rag',
@@ -244,9 +276,18 @@ describe('renderFindingSubject', () => {
       { kind: 'knowledge.raw', sourceId: 'irs-pub-501-2024' },
       { kind: 'knowledge.stale', slug: 'old-vat-rates' },
       { kind: 'system-prompt', section: 'request-classification' },
+      { kind: 'skill', name: 'linear-triage' },
       { kind: 'tool-doc', tool: 'list_invoices' },
       { kind: 'tool-doc', tool: 'list_invoices', aspect: 'examples' },
       { kind: 'new-tool', name: 'diff_csv' },
+      { kind: 'mcp', server: 'linear' },
+      { kind: 'mcp', server: 'linear', tool: 'get_issue' },
+      { kind: 'hook', name: 'post-tool-use' },
+      { kind: 'subagent', name: 'reviewer' },
+      { kind: 'workflow', name: 'linear-ticket' },
+      { kind: 'rollout-policy', field: 'maxTurns' },
+      { kind: 'agent-profile', field: 'model' },
+      { kind: 'code', path: 'src/dispatch.ts' },
       { kind: 'rag', corpus: 'irs-rulings', docId: 'rev-rul-2024-12' },
       { kind: 'memory', key: 'last-customer-id' },
       { kind: 'scaffolding', concern: 'retry-policy' },
@@ -293,4 +334,21 @@ describe('KIND_EXPECTED_SUBJECTS', () => {
     expect(improvement).not.toContain('websearch.outdated')
     expect(improvement).not.toContain('prior-run-summary')
   })
+
+  it.each([
+    ['failure-mode', FAILURE_MODE_KIND_SPEC],
+    ['improvement', IMPROVEMENT_KIND_SPEC],
+    ['knowledge-gap', KNOWLEDGE_GAP_KIND_SPEC],
+    ['knowledge-poisoning', KNOWLEDGE_POISONING_KIND_SPEC],
+  ])('%s actor prompt embeds its complete allowed grammar', (kindId, spec) => {
+    const grammar = findingSubjectGrammarPromptFor(kindId)
+    expect(spec.actorDescription).toContain(grammar)
+    for (const subjectKind of KIND_EXPECTED_SUBJECTS[kindId] ?? []) {
+      expect(grammar).toContain(FINDING_SUBJECT_SYNTAX[subjectKind])
+    }
+  })
+
+  it('rejects an unknown analyst kind instead of emitting an empty grammar', () => {
+    expect(() => findingSubjectGrammarPromptFor('unknown')).toThrow(/unknown analyst kind/)
+  })
 })