fix(plugin): trim SKILL.md description to fit the 1024-char skill spec limit

plugins/kbagent/skills/kbagent/SKILL.md

@@ -1,80 +1,20 @@
 ---
 name: kbagent
 description: >
-  Use when working with Keboola Connection projects via kbagent CLI.
-  Covers: exploring and searching configurations (extractors, writers, transformations),
-  browsing job history, analyzing cross-project data lineage, calling MCP tools
-  across multiple projects, managing development branches, debugging SQL in
-  temporary workspaces, bulk-onboarding organizations, syncing project configs
-  as local files (GitOps), git-branching with Keboola dev branch isolation,
-  sharing buckets across projects, linking shared data,
-  encrypting secrets for MCP tool call workflows,
-  uploading/downloading Storage Files with tag management,
-  and syncing storage metadata and job history. Triggers: kbagent, Keboola project,
-  keboola configs, keboola jobs, keboola lineage, keboola transformations,
-  keboola MCP tools, keboola workspace, SQL debugging, keboola branches,
-  keboola organization, keboola sharing, bucket sharing, link bucket,
-  keboola sync, keboola git,
-  keboola gitops, sync pull, sync push, sync diff, branch-link,
-  search configs, find in configurations, audit configurations,
-  input mapping migration, remove input mapping, Snowflake paths,
-  MULTI_STATEMENT_COUNT, statement count error, SQL transformation migration,
-  keboola encrypt, encrypt secrets, encrypt credentials, encrypt password,
-  keboola encryption API, #password, #api_token, KBC::ProjectSecure,
-  safe config write, dry-run preview, fresh fetch before edit,
-  stale local config file, config version overwrite,
-  default bucket, output bucket, default_bucket, storage.output,
-  raw mode bucket override, custom output bucket name,
-  data app, data apps, keboola data app, streamlit app, streamlit deployment,
-  flask app, fastapi app, node app, python-js, deploy data app,
-  data-app create, data-app deploy, data-app password, data-app start,
-  data-app logs, container logs, app logs, tail logs, build logs,
-  app stdout, app stderr, troubleshoot data app, debug data app,
-  app proxy, simpleAuth, app auto-suspend, configVersion, redeploy contract,
-  Data Science API, /apps endpoint, app password, KBC::Project ciphertext,
-  data-app secrets, app secrets, app runtime secrets, secrets-set,
-  secrets-list, secrets-get, secrets-remove, encrypt app secret,
-  app environment variable, validate repo, validate-repo,
-  data-app golden rule, pre-flight repo check, repo structure check,
-  managed git repo, use-managed-git-repo, Keboola-hosted repo,
-  git-credentials, git-credentials-create, git-repo,
-  deploy from git, managed repository,
-  data app runs, deploy attempts, deploy failure reason, why deploy stopped,
-  local workspace, project directory, kbagent init,
-  invite user, invite member, project invitation, manage members,
-  list members, remove member, change role, project role,
-  bulk invite, invite from CSV, project access, member management,
-  manage token prompt, --allow-env-manage-token, KBC_MANAGE_API_TOKEN,
-  feature flag, feature flags, list features, project features, user features,
-  enable feature, disable feature, set feature flag, add feature, remove feature,
-  early-adopter-preview, direct-access, pay-as-you-go, /manage/features,
-  super admin token, super-admin feature, stack feature catalogue,
-  data stream, data streams, keboola data streams, stream source, OTLP,
-  OpenTelemetry, otel, OTLP endpoint, OTEL_EXPORTER_OTLP_ENDPOINT, telemetry ingest,
-  logs metrics traces, stream create-source, stream detail, stream list,
-  stream delete, otlp source, http source, stream-in, ingest endpoint,
-  scoped token, scoped storage token, token create, token delete, token refresh,
-  mint token, revoke token, rotate token, short-lived token, expiring token,
-  bucket-write token, single-bucket write token, canManageTokens, device enrollment,
-  per-device credential, per-device token, per-device stream source,
-  create_scoped_token, create_stream_source, delete_token, refresh_token,
-  semantic-layer, semantic layer, semantic-layer model, metastore,
-  semantic-metric, semantic-dataset, semantic-relationship,
-  semantic-constraint, semantic-glossary, add metric, edit metric,
-  rename metric, remove metric, validate model, validate semantic layer,
-  promote model, semantic-layer build, semantic-layer export,
-  semantic-layer diff, semantic-layer import, semantic-layer token,
-  metric SQL, dataset FQN, constraint rule, threshold constraint,
-  4-band health, _critical _warning _healthy _review, CODE_METRIC,
-  DIM_METRIC_THRESHOLD, dangling metric FK, orphaned constraint,
-  phantom field, AGG on STRING, SUM on PCT, deep validate,
-  sl, kbagent sl, semantic layer wizard, sl-build, sl-add, sl-edit,
-  reference data, semantic-layer reference-data, chart of accounts, COA,
-  dimension members, account list, dimension data, member list,
-  developer portal, dev-portal, apps-api, register component, vendor app,
-  portal property, ui-options, encryption portal, defaultBucket portal,
-  app icon, configurationSchema portal, publish component, deprecate component,
-  kbagent dev-portal, portal identity, vendor login, service account portal.
+  Use when working with Keboola Connection projects via the kbagent CLI.
+  Covers: exploring and searching component configurations, job history and
+  job runs, cross-project data lineage, Keboola MCP tools, development
+  branches, SQL debugging in temporary workspaces, GitOps sync of configs as
+  local files (pull/push/diff/clone), bucket sharing and linking, encrypting
+  secrets, Storage tables and files, data apps (create/deploy/logs/secrets),
+  conditional flows and schedules, project members and invitations, feature
+  flags, OTLP data streams, scoped Storage tokens, the semantic layer
+  (models, datasets, metrics, constraints, reference data), and the Keboola
+  Developer Portal. Triggers: kbagent, Keboola, keboola config, keboola job,
+  keboola lineage, keboola sync, gitops, dev branch, workspace SQL, data app,
+  streamlit deploy, semantic layer, sl, dev-portal, data stream, OTLP,
+  scoped token, bucket sharing, encrypt secrets, feature flag, flow schedule,
+  invite member.
 ---
 
 # kbagent -- Keboola Agent CLI

tests/test_skill_frontmatter.py

@@ -0,0 +1,55 @@
+"""Static compliance tests for the kbagent skill frontmatter.
+
+Claude Desktop (and the Agent Skills spec) enforce a hard limit of 1024
+characters on the ``description`` field in SKILL.md frontmatter; a longer
+description makes the whole skill fail to load with
+"field 'description' in SKILL.md must be at most 1024 characters"
+(issue #447). These tests keep the frontmatter within every consumer's
+limits so the plugin stays loadable in Claude Desktop, Claude Code, and
+claude.ai alike.
+"""
+
+from __future__ import annotations
+
+import re
+from pathlib import Path
+
+import pytest
+import yaml
+
+REPO_ROOT = Path(__file__).parent.parent
+SKILL_MD = REPO_ROOT / "plugins" / "kbagent" / "skills" / "kbagent" / "SKILL.md"
+
+# Hard limit from the Agent Skills spec, enforced by Claude Desktop at
+# skill-load time (issue #447).
+DESCRIPTION_CHAR_LIMIT = 1024
+
+
+@pytest.fixture(scope="module")
+def frontmatter() -> dict:
+    text = SKILL_MD.read_text(encoding="utf-8")
+    match = re.match(r"^---\n(.*?)\n---\n", text, re.DOTALL)
+    assert match is not None, "SKILL.md is missing YAML frontmatter"
+    data = yaml.safe_load(match.group(1))
+    assert isinstance(data, dict), "SKILL.md frontmatter must be a YAML mapping"
+    return data
+
+
+class TestSkillFrontmatter:
+    def test_skill_file_exists(self) -> None:
+        assert SKILL_MD.is_file(), f"SKILL.md missing at {SKILL_MD}"
+
+    def test_name_matches_directory(self, frontmatter: dict) -> None:
+        assert frontmatter.get("name") == "kbagent"
+
+    def test_description_present(self, frontmatter: dict) -> None:
+        description = frontmatter.get("description", "")
+        assert description.strip(), "SKILL.md description must not be empty"
+
+    def test_description_within_claude_desktop_limit(self, frontmatter: dict) -> None:
+        description = frontmatter["description"]
+        assert len(description) <= DESCRIPTION_CHAR_LIMIT, (
+            f"SKILL.md description is {len(description)} characters; Claude "
+            f"Desktop rejects skills whose description exceeds "
+            f"{DESCRIPTION_CHAR_LIMIT} characters (issue #447). Trim it."
+        )