@automagik/genie 4.260331.1 → 4.260331.2

package/.claude-plugin/marketplace.json

@@ -10,7 +10,7 @@
   "plugins": [
     {
       "name": "genie",
-      "version": "4.260331.1",
+      "version": "4.260331.2",
       "source": "./plugins/genie",
       "description": "Human-AI partnership for Claude Code. Share a terminal, orchestrate workers, evolve together. Brainstorm ideas, wish them into plans, make with parallel agents, ship as one team. A coding genie that grows with your project."
     }

package/.genie/agents/metrics-updater/AGENT.md

@@ -1,199 +1,29 @@
----
-name: metrics-updater
-description: Self-improving daily metrics agent — updates README.md with live project metrics, refines its own prompt, and generates optimization tools
-version: 2
-created: 2026-03-24
-last_refined: null
----
+# metrics-updater Agent
 
-# Metrics Updater Agent
+Fetches live GitHub metrics daily and updates the README metrics table.
 
-## Purpose
+## Metrics Collected
 
-Update README.md with live project metrics daily. After each run, analyze performance, call `/refine` to improve this prompt, and generate tools to optimize future runs. The goal: measurably faster execution every day.
+- **Releases/day** — releases published in the last 24h
+- **Avg merge time** — mean lead time (created→merged) for PRs closed in last 7d
+- **SHIP rate** — merged / total closed PRs in last 7d (%)
+- **Merged PRs (7d)** — count of merged PRs in the rolling 7-day window
 
-## Repository
+## README Markers
 
-- **Owner:** automagik-dev
-- **Repo:** genie
-- **Branch:** dev (metrics commits go here)
+The agent updates the block between `<!-- METRICS:START -->` and `<!-- METRICS:END -->` in README.md.
 
-## Metrics to Fetch
+If markers are missing, insert the table after the badges block, before `## What is Genie?`.
 
-| Metric | Source | Calculation |
-|--------|--------|-------------|
-| Releases/day | `gh api repos/{owner}/{repo}/releases` | Count releases created in last 24h |
-| Avg bug-fix time | `gh api repos/{owner}/{repo}/pulls?state=closed` | Mean time from PR open → merge for bug-fix PRs (last 7 days) |
-| SHIP rate | `gh api repos/{owner}/{repo}/pulls?state=closed` | % of PRs that shipped without FIX-FIRST (merged on first review cycle) |
-| Lines changed (24h) | `git log --since="24 hours ago" --stat` | Total insertions + deletions in the last 24 hours |
-| Commits (24h) | `git log --since="24 hours ago" --oneline` | Count of commits in the last 24 hours |
-| Pull requests (24h) | `gh api search/issues` | Count of PRs created in the last 24 hours |
-| Parallel agents | `genie status` or process count | Number of active genie workers at time of run |
+## Commit Convention
 
-## Execution Steps
-
-Run these steps in order. Time each step for performance tracking.
-
-### Phase 1: Metrics Update (run-metrics.sh)
-
-1. **Load state** — Read `state.json` for last_metrics (fallback if API fails)
-2. **Fetch metrics** — Call GitHub API via `gh` CLI for each metric
-   - Use `tools/cached-fetch.sh` if available (avoids redundant API calls)
-3. **Calculate** — Parse API responses, compute aggregated numbers
-   - Use `tools/fast-parse.py` if available (single-pass optimization)
-4. **Update README** — Find or create metrics table in README.md, update values
-5. **Commit** — `chore: update live metrics (X/day, Yh avg, Z% SHIP)`
-   - Use `tools/batch-commit.sh` if available (batched git operations)
-6. **Log run** — Append structured JSON to `runs.jsonl` with step timings
-
-```bash
-# Full run (fetch, update README, commit, log):
-bash tools/run-metrics.sh
-
-# Dry run (no commit):
-bash tools/run-metrics.sh --dry-run
 ```
-
-### Phase 2: Self-Refinement
-
-After the metrics update completes:
-
-7. **Analyze performance** — Run perf-analyzer to identify bottlenecks
-   ```bash
-   python3 tools/perf-analyzer.py --format text
-   ```
-
-8. **Generate tools** — Create optimization tools for slow steps
-   ```bash
-   python3 tools/generate-tools.py
-   ```
-
-9. **Prepare refinement context** — Build context from performance data
-   ```bash
-   bash tools/self-refine.sh
-   ```
-
-10. **Refine prompt** — Call `/refine` in file mode to optimize this prompt
-    ```
-    /refine @.genie/agents/metrics-updater/AGENT.md
-    ```
-
-11. **Verify** — Confirm AGENT.md was updated and state.json has `last_refined_at`
-
-## README Metrics Table Format
-
-Insert after the badges block, before "## What is Genie?". The table uses HTML comment signature markers with an ISO 8601 timestamp — no "Updated" column needed.
-
-```markdown
-<!-- METRICS:START — Updated by Genie Metrics Agent at 2026-03-24T23:45:00Z -->
-| Metric | Value |
-|--------|-------|
-| Releases/day | 17 |
-| Avg bug-fix time | 1.7h |
-| SHIP rate | 100% |
-| Lines changed (24h) | 12,450 |
-| Commits (24h) | 34 |
-| Pull requests (24h) | 8 |
-| Parallel agents | 5 |
-<!-- METRICS:END — 🧞 automagik/genie -->
+chore: update live metrics (X/day, Yh avg, Z% SHIP)
 ```
 
-### Quality Expectations
-
-- The table MUST have exactly 7 metric rows — no more, no less
-- Values MUST be real data from GitHub API and git log — never hardcoded placeholders
-- The ISO timestamp in the START marker MUST reflect the actual update time in UTC
-- The bottom marker includes the Genie signature (`🧞 automagik/genie`) for attribution
-- Numbers should be human-readable: use comma separators for LoC (e.g., `12,450`)
-
-## Tools Available
-
-Source tools from `tools/` directory before executing:
-
-### Core Tools (Wave 1 — always present)
-- `tools/run-metrics.sh` — **Main orchestrator** (fetch → parse → update README → commit → log with step timing)
-- `tools/github-api.sh` — GitHub API wrapper with caching and retry
-- `tools/parse-metrics.py` — Metrics parser and calculator
-- `tools/update-readme.py` — README metrics table updater (finds METRICS:START/END markers)
-- `tools/commit-formatter.sh` — Clean commit message formatter
-
-### Self-Improvement Tools (Wave 2 — refinement loop)
-- `tools/perf-analyzer.py` — Analyzes runs.jsonl for bottlenecks, trends, and optimization recommendations
-- `tools/self-refine.sh` — Prepares refinement context and triggers `/refine` on AGENT.md
-- `tools/generate-tools.py` — Analyzes perf data and generates optimization tools for slow steps
-
-### Auto-Generated Tools (created by generate-tools.py)
-- `tools/cached-fetch.sh` — Cached GitHub API fetcher with TTL (generated when fetch steps are slow)
-- `tools/fast-parse.py` — Optimized single-pass metrics parser (generated when parse steps are slow)
-- `tools/batch-commit.sh` — Batched git operations (generated when commit steps are slow)
-
-New tools may be generated after each run. Check `tools/` for the latest inventory.
-
-## Constraints
-
-- **MUST** call `/refine` after each run with performance data
-- **MUST** log every run to `runs.jsonl` with: timestamp, duration_ms, api_calls, tools_generated, errors, steps, slowest_step
-- **MUST** fall back to `state.json` last_metrics if GitHub API fails
-- **MUST NOT** push directly to main — commit to dev branch only
-- **MUST NOT** run more than once per day
-- **SHOULD** generate new tools when identifying slow operations (>500ms average)
-- **SHOULD** use cached API responses when available (tools/cached-fetch.sh)
-- **SHOULD** use fast-parse.py when available for metrics calculation
-
-## Performance Tracking
-
-After each run, `runs.jsonl` captures:
-- `timestamp` — When the run started
-- `duration_ms` — Total execution time
-- `api_calls` — Number of GitHub API calls
-- `tools_generated` — Tools created this run
-- `tools_available` — Total tools in tools/ directory
-- `errors` — Array of error messages
-- `status` — success | no_changes | failed
-- `fallback` — Whether fallback metrics were used
-- `slowest_step` — Name of the step that took longest
-- `steps` — Array of `{name, duration_ms}` for each execution step
-- `metrics` — The computed metrics values
-
-### Step Names (for performance analysis)
-- `load_state` — Reading state.json
-- `fetch_releases` — GitHub API call for releases
-- `fetch_prs` — GitHub API call for PRs
-- `count_agents` — Counting parallel agents
-- `parse_metrics` — Computing metrics from API data
-- `update_readme` — Updating README.md
-- `update_state` — Writing state.json
-- `commit` — Git add + commit
-
-## Self-Refinement Protocol
-
-After completing the metrics update (Phase 1), execute Phase 2:
-
-1. **Run perf-analyzer.py** to get a performance report with bottleneck analysis
-2. **Run generate-tools.py** to create optimization tools for slow steps
-3. **Run self-refine.sh** to prepare refinement context and append it to AGENT.md
-4. **Call `/refine @AGENT.md`** to optimize the prompt based on performance data
-5. **Verify** the refined prompt preserves:
-   - Core metrics (releases/day, bug-fix time, SHIP rate, parallel agents)
-   - The self-refinement protocol section
-   - Tool references and execution steps
-   - Version number incremented in frontmatter
-
-The refined prompt replaces this file for the next run. Tools persist in `tools/`.
-
-## Graceful Degradation
-
-If GitHub API is unavailable:
-1. Read `state.json` for `last_metrics`
-2. Use yesterday's values in README (do not update the timestamp in the START marker)
-3. Log the error to `runs.jsonl` with step timings
-4. Skip commit (no changes to README)
-5. Still call `/refine` to analyze the failure and improve error handling
+Push to `dev` branch only. Never push to `main`.
 
-## Improvement Targets
+## State Files
 
-Track these across runs (visible in perf-analyzer.py output):
-- **Day 1 → Day 7 execution time**: Target 50% reduction
-- **API calls per run**: Target ≤2 (with caching)
-- **Tools generated**: Target ≥3 by Day 2
-- **Error rate**: Target 0% after initial stabilization
+- `state.json` — last successful metrics run
+- `runs.jsonl` — append-only log of every run

package/.genie/agents/metrics-updater/runs.jsonl

@@ -12,3 +12,4 @@
 {"timestamp":"2026-03-28T00:00:00Z","duration_ms":45000,"api_calls":3,"tools_generated":0,"errors":[],"metrics":{"releases_24h":0,"avg_merge_time_hours":2.9,"ship_rate_pct":94,"merged_prs_7d":47}}
 {"timestamp":"2026-03-29T00:00:00Z","duration_ms":52000,"api_calls":2,"tools_generated":0,"errors":[],"metrics":{"releases_24h":0,"avg_merge_time_hours":3.0,"ship_rate_pct":92,"merged_prs_7d":46}}
 {"timestamp":"2026-03-30T00:00:00Z","duration_ms":5000,"api_calls":3,"tools_generated":0,"errors":0,"metrics":{"releases_24h":0,"releases_per_day":0,"avg_merge_time_hours":0.3,"ship_rate_pct":84,"merged_prs_7d":32,"closed_prs_7d":38}}
+{"timestamp":"2026-03-31T00:00:00Z","duration_ms":5002,"api_calls":2,"tools_generated":4,"errors":0,"metrics":{"releases_24h":0,"merged_prs_7d":35,"avg_merge_time_h":0.3,"ship_rate_pct":85}}

package/.genie/agents/metrics-updater/state.json

@@ -1,11 +1,9 @@
 {
-  "last_run": "2026-03-30T00:00:00Z",
+  "last_run": "2026-03-31T00:00:00Z",
   "last_metrics": {
     "releases_24h": 0,
-    "releases_per_day": 0,
-    "avg_merge_time_hours": 0.3,
-    "ship_rate_pct": 84,
-    "merged_prs_7d": 32,
-    "closed_prs_7d": 38
+    "merged_prs_7d": 35,
+    "avg_merge_time_h": 0.3,
+    "ship_rate_pct": 85
   }
 }

package/README.md

@@ -17,10 +17,10 @@
 
 | Metric | Value | Updated |
 |--------|-------|---------|
-| Releases/day | **0** | 2026-03-30 |
-| Avg merge time | **0.3h** | 2026-03-30 |
-| SHIP rate | **84%** | 2026-03-30 |
-| Merged PRs (7d) | **32** | 2026-03-30 |
+| Releases/day | **0** | 2026-03-31 |
+| Avg merge time | **0.3h** | 2026-03-31 |
+| SHIP rate | **85%** | 2026-03-31 |
+| Merged PRs (7d) | **35** | 2026-03-31 |
 
 <!-- METRICS:END -->
 

package/openclaw.plugin.json

@@ -2,7 +2,7 @@
   "id": "genie",
   "name": "Genie",
   "description": "Skills, agents, and hooks for the Genie CLI terminal orchestration toolkit",
-  "version": "4.260331.1",
+  "version": "4.260331.2",
   "configSchema": {
     "type": "object",
     "additionalProperties": false,

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@automagik/genie",
-  "version": "4.260331.1",
+  "version": "4.260331.2",
   "description": "Collaborative terminal toolkit for human + AI workflows",
   "type": "module",
   "bin": {

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

@@ -1,6 +1,6 @@
 {
   "name": "genie",
-  "version": "4.260331.1",
+  "version": "4.260331.2",
   "description": "Human-AI partnership for Claude Code. Share a terminal, orchestrate workers, evolve together. Brainstorm ideas, turn them into wishes, execute with /work, validate with /review, and ship as one team.",
   "author": {
     "name": "Namastex Labs"

package/plugins/genie/package.json

@@ -1,6 +1,6 @@
 {
   "name": "genie-plugin",
-  "version": "4.260331.1",
+  "version": "4.260331.2",
   "private": true,
   "description": "Runtime dependencies for genie bundled CLIs",
   "type": "module",