@onlooker-community/ecosystem 0.15.0 → 0.15.2

package/.claude-plugin/plugin.json

@@ -1,6 +1,6 @@
 {
   "name": "ecosystem",
-  "version": "0.15.0",
+  "version": "0.15.2",
   "description": "Observability substrate for Claude Code. Provides the shared ~/.onlooker/ storage root, canonical schema-validated event emission, session and tool tracking hooks, and prompt rules. Required by all other Onlooker plugins.",
   "author": {
     "name": "Onlooker Community",

package/.github/workflows/release.yml

@@ -33,19 +33,23 @@ jobs:
       # Code marketplace, not npm — a release for a sibling alone must not
       # trigger `npm publish`, which would attempt to re-publish ecosystem at
       # its existing version and fail.
+      #
+      # We check paths_released (a JSON array) rather than the per-path output
+      # key `.--release_created` because keys starting with `.` are silently
+      # dropped when release-please-action writes them to $GITHUB_OUTPUT.
       - uses: actions/checkout@v4
-        if: ${{ steps.release.outputs['.--release_created'] }}
+        if: ${{ contains(fromJSON(steps.release.outputs.paths_released || '[]'), '.') }}
 
       - uses: actions/setup-node@v4
-        if: ${{ steps.release.outputs['.--release_created'] }}
+        if: ${{ contains(fromJSON(steps.release.outputs.paths_released || '[]'), '.') }}
         with:
           node-version: '22'
           registry-url: 'https://registry.npmjs.org'
 
       - run: npm ci
-        if: ${{ steps.release.outputs['.--release_created'] }}
+        if: ${{ contains(fromJSON(steps.release.outputs.paths_released || '[]'), '.') }}
 
       - run: npm publish
-        if: ${{ steps.release.outputs['.--release_created'] }}
+        if: ${{ contains(fromJSON(steps.release.outputs.paths_released || '[]'), '.') }}
         env:
           NODE_AUTH_TOKEN: ${{ secrets.NPM_TOKEN }}

package/.release-please-manifest.json

@@ -1,7 +1,7 @@
 {
-  ".": "0.15.0",
+  ".": "0.15.2",
   "plugins/archivist": "0.1.0",
-  "plugins/tribunal": "1.0.0",
+  "plugins/tribunal": "1.0.1",
   "plugins/echo": "0.2.0",
   "plugins/cartographer": "0.2.0"
 }

package/CHANGELOG.md

@@ -1,11 +1,18 @@
-# [0.8.0](https://github.com/onlooker-community/ecosystem/compare/v0.7.2...v0.8.0) (2026-05-22)
+# Changelog
 
+## [0.15.2](https://github.com/onlooker-community/ecosystem/compare/ecosystem-v0.15.1...ecosystem-v0.15.2) (2026-05-25)
 
-### Features
 
-* **hooks:** add TaskCreated and TaskCompleted task lifecycle trackers ([#21](https://github.com/onlooker-community/ecosystem/issues/21)) ([986ffa8](https://github.com/onlooker-community/ecosystem/commit/986ffa84bdd857a464ca0d556671628190ed27bc))
+### Bug Fixes
 
-# Changelog
+* **tribunal:** persist all artifacts on every iteration including retries :relieved: ([#41](https://github.com/onlooker-community/ecosystem/issues/41)) ([1636105](https://github.com/onlooker-community/ecosystem/commit/163610535a4ce0fa73c8fb82dc5c6296d2d1065a))
+
+## [0.15.1](https://github.com/onlooker-community/ecosystem/compare/ecosystem-v0.15.0...ecosystem-v0.15.1) (2026-05-25)
+
+
+### Bug Fixes
+
+* **ci:** use paths_released to gate npm publish :rage: ([#37](https://github.com/onlooker-community/ecosystem/issues/37)) ([c62b17f](https://github.com/onlooker-community/ecosystem/commit/c62b17f7e1352cfe260a23c8f48be30f72edbbed))
 
 ## [0.15.0](https://github.com/onlooker-community/ecosystem/compare/ecosystem-v0.14.0...ecosystem-v0.15.0) (2026-05-25)
 

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@onlooker-community/ecosystem",
-  "version": "0.15.0",
+  "version": "0.15.2",
   "description": "Agents, skills, hooks, commands, rules, and MCP configurations that power [Onlooker](https://onlooker.dev)",
   "author": {
     "name": "Onlooker Community",

package/plugins/tribunal/.claude-plugin/plugin.json

@@ -1,6 +1,6 @@
 {
   "name": "tribunal",
-  "version": "1.0.0",
+  "version": "1.0.1",
   "description": "Multi-agent execution with LLM-as-a-Judge quality gates. An Actor performs work; a jury of typed Judges scores it against a project-overridable rubric; a Meta-Judge reviews the jury for bias; the gate decides accept, retry, or exhaust. Grounded in LLM-as-a-Judge (Zheng et al. 2023) and LLM-as-a-Meta-Judge (Wu et al. 2024). Builds on the Onlooker ecosystem plugin.",
   "author": {
     "name": "Onlooker Community",

package/plugins/tribunal/CHANGELOG.md

@@ -1,5 +1,12 @@
 # Changelog
 
+## [1.0.1](https://github.com/onlooker-community/ecosystem/compare/tribunal-v1.0.0...tribunal-v1.0.1) (2026-05-25)
+
+
+### Bug Fixes
+
+* **tribunal:** persist all artifacts on every iteration including retries :relieved: ([#41](https://github.com/onlooker-community/ecosystem/issues/41)) ([1636105](https://github.com/onlooker-community/ecosystem/commit/163610535a4ce0fa73c8fb82dc5c6296d2d1065a))
+
 ## 1.0.0 (2026-05-24)
 
 

package/plugins/tribunal/skills/tribunal/SKILL.md

@@ -50,14 +50,22 @@ Emit `tribunal.session.start` with the resolved config (`judge_types`, `gate_pol
 
 For `iteration_number` from `0` while `iteration_number < max_iterations`:
 
-1. **Iteration start.** Generate `iteration_id=$(tribunal_ulid)`. `trigger` is `"initial"` for n=0, `"gate_blocked"` for retries. Emit `tribunal.iteration.start`.
+1. **Iteration start.** Generate `iteration_id=$(tribunal_ulid)`. `trigger` is `"initial"` for n=0, `"gate_blocked"` for retries. Initialize the iteration directory (creates `verdicts/` subdirectory):
+   ```bash
+   tribunal_init_iteration "$project_key" "$task_id" "$iteration_id"
+   ```
+   Emit `tribunal.iteration.start`.
 
 2. **Actor.** Emit `tribunal.actor.start`. Use the Task tool to spawn `tribunal-actor` with:
    - The task description.
    - The rubric criteria (just `name` + `weight` + `min_pass`).
    - On retries: a digest of the prior iteration's consensus, dissent (if any), and Meta-Judge override.
 
-   Capture the Actor's final output. Persist it: `tribunal_write_actor_output "$project_key" "$task_id" "$iteration_id" "$actor_output"`. Emit `tribunal.actor.complete` with `success: true` and the inferred `artifact_kind` (`file` / `patch` / `message` / `command`).
+   Capture the Actor's final output. **`$actor_output` must be the verbatim, complete text returned by the Task tool — never a summary, paraphrase, or placeholder string.** Persist it:
+   ```bash
+   tribunal_write_actor_output "$project_key" "$task_id" "$iteration_id" "$actor_output"
+   ```
+   Emit `tribunal.actor.complete` with `success: true` and the inferred `artifact_kind` (`file` / `patch` / `message` / `command`).
 
 3. **Empanel the jury.** Resolve the panel from configured types:
    ```bash
@@ -67,17 +75,31 @@ For `iteration_number` from `0` while `iteration_number < max_iterations`:
    [[ -n "$rubric_types" && "$rubric_types" != "null" ]] && types="$rubric_types"
    jury=$(tribunal_jury_empanel "$types")
    ```
-   Persist the jury (`tribunal_write_iteration_artifact ... jury ...`) and emit `tribunal.jury.empaneled` with the schema-shaped `judges[]` (`tribunal_jury_to_schema_judges "$jury"`).
+   Persist the jury and emit `tribunal.jury.empaneled`:
+   ```bash
+   tribunal_write_iteration_artifact "$project_key" "$task_id" "$iteration_id" "jury" "$jury"
+   schema_judges=$(tribunal_jury_to_schema_judges "$jury")  # pass $schema_judges as judges[] in the event
+   ```
 
 4. **Run each Judge.** For each entry in the jury panel:
    - Emit `tribunal.judge.start` with `judge_id`, `judge_type`, `judge_model_id`.
    - Spawn the judge subagent (`.subagent` field) with the Actor output + rubric.
    - Parse the JSON object the judge returns. Augment it with `task_id`, `iteration_id`, `judge_id`, `judge_model_id` from the panel entry, and `judge_type` from the panel entry (canonical, overriding what the agent self-reported).
    - Emit `tribunal.verdict` with that payload.
-   - Persist with `tribunal_write_judge_verdict`.
+   - **Persist the verdict. This call is required for every judge on every iteration — including retries:**
+     ```bash
+     tribunal_write_judge_verdict "$project_key" "$task_id" "$iteration_id" "$judge_id" "$verdict_json"
+     ```
 
    Collect the verdicts into a JSON array `verdicts`.
 
+   **Before moving to step 5, verify all artifacts written so far are on disk:**
+   - `iteration-<id>/actor.md` — verbatim actor output (written in step 2)
+   - `iteration-<id>/jury.json` — jury panel (written in step 3)
+   - `iteration-<id>/verdicts/<judge_id>.json` — one file per judge (written in step 4)
+
+   (`iteration-<id>/gate.json` is written in step 7 — verify it there.)
+
 5. **Aggregate + dissent.**
    ```bash
    method=$(printf '%s' "$rubric" | jq -r '.aggregation_method // "weighted_mean"')