chore: version packages

.claude-plugin/plugin.json

@@ -1,7 +1,7 @@
 {
   "name": "taskless",
   "description": "Taskless skills for code quality rules, authentication, and project management",
-  "version": "0.13.4",
+  "version": "0.13.5",
   "author": {
     "name": "Taskless"
   },

package.json

@@ -1,7 +1,7 @@
 {
   "private": true,
   "name": "@taskless/skills",
-  "version": "0.13.4",
+  "version": "0.13.5",
   "license": "MIT",
   "repository": "taskless/cli.git",
   "scripts": {

packages/cli/CHANGELOG.md

@@ -1,5 +1,16 @@
 # @taskless/cli
 
+## 0.13.5
+
+[Compare with v0.13.4](https://github.com/taskless/cli/compare/v0.13.4...v0.13.5)
+
+### Patch Changes
+
+- ef2744a: Reconcile now tells Taskless whether it is running in CI. Every `POST /cli/api/v2/reconcile` carries `environment: { ci, provider? }`, using the CI detection telemetry already performs (`provider` is the hyphenated name, such as `github-actions`, when it is recognized). The service uses it only to record when a repository last reconciled from CI, which drives the dashboard's "CI set up" indicator; verdicts, entitlements and output are unchanged, and a local run sends `ci: false`.
+- 0b41364: The `no-eval-call` entry in `@taskless/cli/reference.json` now flags a call to `eval` with any number of arguments, as its prompt asks, and carries a fail case for `eval()` and for `eval(a, b)`. Previously the reference rule matched exactly one argument, so a rule generated correctly from the prompt, with fixtures on each side of the arity boundary as `create-sg-rule` recommends, was graded against a reference that did not fire on them. The prompt now names the arities it expects. The corpus `version` is unchanged: the shape is the same, only the content of one entry moved.
+- 0b2ed45: The `prefer-use-over-utilize` entry in `@taskless/cli/reference.json` now flags `utilized` and `utilization`, as its prompt asks, and carries a fail case for each. Previously the reference rule missed both, so a rule generated correctly from the prompt was graded against a reference that did not fire on those cases. The prompt now names the forms it expects. The corpus `version` is unchanged: the shape is the same, only the content of one entry moved.
+- 333cc04: `taskless verify --json` now reports every fixture case under each rule's new `cases` array, for `sg`, `vale` and `runtime` alike: `{ name, expected, ok, message?, files? }`, where `name` is the case's path under `tests/` (such as `fail/cast`). A failing case carries the same line `errors` prints for it as `message`, and its files as `files` (project-relative paths, the spelling `findings[].file` uses; a file over 64 KiB or one that is not text is left out). The entries are built from each runner's own per-case verdict, so a consumer no longer has to parse `errors` to learn which case failed. The field is additive: `errors` is unchanged, and `cases` is always present, empty when no case ran (for `validate`, a refused runtime run, or a rule that failed before its fixtures ran). It is also declared in the schema `@taskless/cli/schemas` publishes.
+
 ## 0.13.4
 
 [Compare with v0.13.3](https://github.com/taskless/cli/compare/v0.13.3...v0.13.4)

packages/cli/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@taskless/cli",
-  "version": "0.13.4",
+  "version": "0.13.5",
   "license": "MIT",
   "repository": {
     "type": "git",

skills/taskless/SKILL.md

@@ -22,7 +22,7 @@ description: |
   `agent route`; it does NOT suppress the skill.
 metadata:
   author: taskless
-  version: 0.13.4
+  version: 0.13.5
 compatibility: Designed for Agents implementing the Agent Skills specification.
 ---