pantheon-opencode 1.1.1 → 1.1.3-beta.0

package/commands/pantheon-audit.md

@@ -0,0 +1,25 @@
+---
+description: "Audit: heuristic scan (zero LLM) + Themis review. Usage: /pantheon-audit [--light|--full|--plan]"
+agent: "zeus"
+---
+# /pantheon-audit — Quality & Security Audit
+
+**What:** Layer 1 heuristic scan (zero tokens, <2s). If BLOCKING, returns score. If APPROVED, Themis deep review.
+
+**Usage:**
+- `/pantheon-audit` — Full audit (Layer 1 → 2 → 3)
+- `/pantheon-audit --light` — Layer 1 only (quick scan)
+- `/pantheon-audit --full` — Force full review
+- `/pantheon-audit --plan` — Verification planning only
+
+## Layers
+
+### Layer 1 — Heuristic Scan
+`python3 scripts/themis_heuristic_scan.py [--path=<dir>]`
+Zero tokens, <2s. Checks: ruff, biome, anti-pattern slop, anti-overengineering (YAGNI), hash-anchored edits. Returns score 0-100 + APPROVED/BLOCKING.
+
+### Layer 2 — Themis Deep Review
+~500 tokens. Confidence scoring, regression prediction, OWASP Top 10.
+
+### Layer 3 — Verification Planning
+For N>5 file changes. Sugere plano de verificação antes de implementar.

package/commands/pantheon-consolidate.md

@@ -0,0 +1,11 @@
+---
+description: "Merge and consolidate memory entries"
+agent: zeus
+---
+
+Merge related memory entries and consolidate duplicates:
+
+1. Search for related memories via memory_search
+2. Merge overlapping entries
+3. Remove duplicates via memory_forget
+4. Store consolidated entry via memory_store

package/commands/pantheon-deepwork.md

@@ -0,0 +1,79 @@
+---
+description: "Start a heavy multi-phase task with persisted checkpoints, phased specialist dispatch, and Themis review gates. Progress saved to .pantheon/deepwork/ for resumability."
+agent: "zeus"
+---
+# /pantheon-deepwork — Heavy Task Workflow
+
+Structured, checkpointed workflow for complex multi-step tasks spanning multiple agents and sessions. Progress persisted to `.pantheon/deepwork/<task-slug>/`. Each phase gated by Themis.
+
+| Use When | Don't Use When |
+|---|---|
+| 10+ turns or multi-session | Simple fixes (→ @talos) |
+| Multi-agent dependencies | Single-file changes (→ /subtask) |
+| Costly context loss | < 5 turns (→ normal delegation) |
+
+## Workflow
+
+```
+Phase 0: SCOPING
+  └─ Zeus + Athena define scope, phases, acceptance criteria → PLAN.md
+
+Phase 1: DISCOVERY
+  └─ @apollo maps relevant codebase areas → DISCOVERY.md
+
+Phase 2-N: IMPLEMENTATION (parallel per phase)
+  └─ @hermes, @aphrodite, @demeter, @hephaestus, @prometheus
+  └─ Output: IMPL-phase-<N>-<agent>.md
+
+GATE after each phase:
+  └─ @themis reviews output → FAIL (fix + resubmit) or PASS (continue)
+
+Phase FINAL: VERIFICATION
+  └─ @themis full integration review → REVIEW.md
+  └─ All tests pass, coverage >80%
+```
+
+## Checkpoints
+
+```
+Usage:  --resume <slug> | --status <slug> | --list
+
+.pantheon/deepwork/<slug>/
+├── PLAN.md  DISCOVERY.md  IMPL-phase-*-*.md  phase-*-review.md  REVIEW.md  STATUS.md
+```
+
+## Full-Auto Mode (`--full-auto`)
+
+> ⚠️ Bypasses human review gates. Tier 1 gates (plan, commit, deploy) still require approval.
+
+**Themis auto-approves non-blocking reviews** — only stops on **BLOCKING** verdict. No waiting between phases. Checkpoints saved at every boundary.
+
+```
+/pantheon-deepwork --full-auto "Refactor auth service to use JWT"
+/pantheon-deepwork --full-auto --resume auth-jwt
+```
+
+## Anti-Stall
+
+- **Stall Detection** — 3 turns no progress → escalate
+- **Phase Reminder** — continue only independent work after dispatch
+- **Progress Checkpoint** — every 5 turns, summarize completed vs remaining
+- **Delegate Retry** — failures retried once with rephrased prompt
+
+## Usage
+
+```
+/pantheon-deepwork "Add user authentication with OAuth2"           # Start new session
+/pantheon-deepwork --resume auth-oauth2                            # Resume interrupted
+/pantheon-deepwork --status auth-oauth2                            # Show progress
+/pantheon-deepwork --list    /pantheon-deepwork --archive <slug>    # List / archive
+/pantheon-deepwork --full-auto "Refactor auth"                     # Themis auto-approve
+/pantheon-deepwork --full-auto --resume auth-refactor              # Resume full-auto
+```
+
+## Safety
+
+- All progress persisted — work is never lost
+- Each phase gated by Themis — quality enforced at every step
+- Explicit resume required — no auto-continue without intent (unless `--full-auto`)
+- `.pantheon/deepwork/` is gitignored — no accidental commits

package/commands/pantheon-optimize.md

@@ -0,0 +1,79 @@
+---
+description: "Scan, clean, cache, and archive Pantheon project files. Scans for bloat, archives completed deepwork, migrates flat files to MCP memory, and reports optimization opportunities. Usage: /pantheon-optimize"
+agent: "zeus"
+---
+# /pantheon-optimize — Pantheon Project Optimizer
+
+**What:** Multi-mode optimization: scans for documentation bloat, archives completed deepwork, migrates memory-bank flat files to MCP memory (ChromaDB), and reports token savings.
+
+## Usage
+
+```
+/pantheon-optimize                          → Full scan + compress (default)
+/pantheon-optimize --dry-run                → Preview only, no changes
+/pantheon-optimize --archive                → Archive completed deepwork sessions
+/pantheon-optimize --cache                  → Migrate memory-bank to MCP memory
+/pantheon-optimize --path=<dir>             → Limit scan to directory
+/pantheon-optimize --all                    → Run all modes (default)
+```
+
+## Modes
+
+### `--archive` — Deepwork Archiving
+Archives completed deepwork sessions from `.pantheon/deepwork/<slug>/` to `.pantheon/deepwork/archive/<slug>/`.
+
+**Criteria:** A session is "completed" if it has a `REVIEW.md` containing "APPROVED".
+
+```
+/pantheon-optimize archive                          → Interactive (ask per slug)
+/pantheon-optimize archive --dry-run                → Preview only
+/pantheon-optimize archive --auto                   → Archive all completed
+/pantheon-optimize archive --auto --keep=2           → Keep only PLAN.md + REVIEW.md
+/pantheon-optimize archive --auto --compress         → Single .md per slug
+```
+
+### `--cache` — Memory Bank Migration
+Scans `.pantheon/memory-bank/` for flat `.md` files and migrates them to the MCP memory server (ChromaDB).
+
+**What it does:**
+1. Reads all `.md` files in `.pantheon/memory-bank/`
+2. Chunks by heading (## sections)
+3. Calls `memory_store()` for each chunk with appropriate namespace
+4. Optionally archives the original files after successful migration
+
+```
+/pantheon-optimize cache                     → Preview candidates, ask confirm
+/pantheon-optimize cache --dry-run           → Show what would migrate
+/pantheon-optimize cache --auto              → Migrate + archive originals
+```
+
+### `--dry-run` (global flag)
+Prevents any destructive action. Shows what WOULD be done.
+
+### Default mode (no flags)
+Runs bloat scan on all `.md` and `.json` files + compression report.
+
+## Output
+
+```
+🧹 Pantheon Optimize Report
+
+📦 Bloat Scan: 47 .md + 12 .json files
+   → 3 oversized files found
+   → ~2.4K tokens recoverable
+
+📁 Deepwork Archive: 15 candidates
+   → 12 completed (REVIEW.md APPROVED)
+   → 3 in progress (skipped)
+   → Run with --archive --auto to archive 12 sessions
+
+🗄️  Memory Cache: 24 files in memory-bank/
+   → 18 already in MCP memory (dedup)
+   → 6 candidates for migration (~3.2K tokens)
+   → Run with --cache to migrate
+
+📊 Summary:
+   ⚡ 15 files optimizable → ~5.6K tokens
+   🗑️  12 deepworks archivable → ~320 files cleanup
+   🗄️  6 memory files → MCP migration ready
+```

package/commands/pantheon.md

@@ -0,0 +1,49 @@
+---
+description: "Dispatch a question to 2-4 specialist agents via inline Zeus council"
+agent: zeus
+---
+# /pantheon — Council Synthesis
+
+**What:** Dispatches the question to 2-3 specialist agents **in parallel** via `task()`, then synthesizes their responses into a single recommendation with resolved divergences.
+**Usage:** `/pantheon <question>`
+**When:** Architecture trade-offs, technology selection, security assessments, multi-stakeholder concerns
+**Returns:** Structured synthesis with recommendation, confidence level, resolved divergences
+
+## Dispatch Selection
+
+Choose 2-3 agents based on domain:
+
+| Domain | Specialists |
+|--------|-------------|
+| Architecture | hermes, demeter, themis, athena |
+| Security | themis, hermes, prometheus, nyx |
+| Database | demeter, hermes, prometheus |
+| AI/RAG | hephaestus, nyx |
+| Infrastructure | prometheus, hermes, themis |
+| Frontend/UX | aphrodite, themis, hermes |
+| Observability | nyx, hermes |
+| General | athena, themis, hermes |
+
+Send ALL `task()` calls in a single message. Each must return: Recommendation · Reasoning · Trade-offs · Risks · Confidence.
+
+## Synthesis Structure
+
+After ALL responses arrive, produce EXACTLY this structure:
+
+### 📋 Individual Assessments
+For each specialist: Recommendation, Reasoning, Trade-offs, Risks, Confidence
+
+### ✅ Agreements
+Where 2+ specialists converge
+
+### ⚡ Divergences
+Where they disagree — resolve each one with reasoning
+
+### 🔥 Total Divergence
+If ALL specialists disagree on a point, call this out explicitly
+
+### 🏆 Recommendation
+Decisive conclusion with confidence level
+
+## Question:
+$ARGUMENTS

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "pantheon-opencode",
-  "version": "1.1.1",
+  "version": "1.1.3-beta.0",
   "description": "Multi-agent orchestration for OpenCode — 14 specialized agents, TDD, Themis review gates.",
   "publishConfig": {
     "access": "public",

package/commands/pantheon-reflect.md

@@ -1,16 +0,0 @@
----
-description: "Reflect on workflow friction and create improvement action items"
-agent: zeus
----
-# /pantheon-reflect — Workflow Reflection
-
-**What:** After completing a task, reflect on what could be improved. Turns repeated friction into reusable skills, agents, or config improvements.
-**Usage:** `/pantheon-reflect [session-or-task]`
-**Returns:** Structured reflection with action items
-
-## When to Use
-- After completing a complex feature
-- When you notice the same manual steps repeating
-- At end of work session
-
-Delegates to @mnemosyne with `skill: reflect`.

package/commands/pantheon-verify.md

@@ -1,17 +0,0 @@
----
-description: "Create verification plan before non-trivial changes"
-agent: zeus
----
-# /pantheon-verify — Verification Planning
-
-**What:** Plans an evidence path before non-trivial changes, defining how to verify correctness.
-**Usage:** `/pantheon-verify <change-description>`
-**Returns:** Verification plan with strategy, rollback, and acceptance criteria
-
-## When to Use
-- Before changes touching 3+ files
-- Before DB migrations or schema changes
-- Before API contract changes
-- Before dependency upgrades
-
-Delegates to @themis + @athena with `skill: verification-planning`.