pantheon-opencode 1.5.0-beta.14 → 1.5.0-beta.15

package/scripts/test-context-compression.sh

@@ -1,169 +0,0 @@
-#!/bin/bash
-# Test: Context Compression Pipeline
-# Simulates Themis APPROVED → triggers compress_context handoff
-#
-# Usage: bash scripts/test-context-compression.sh
-#
-# This script validates the prerequisites for the Level 2 context compression
-# pipeline and creates mock artifacts for end-to-end testing.
-
-set -euo pipefail
-
-echo "=== Context Compression Test ==="
-echo ""
-
-# Step 1: Verify prerequisites
-echo "[1/4] Checking prerequisites..."
-errors=0
-
-test -f skills/context-compression/SKILL.md || { echo "   ❌ SKILL.md not found"; errors=$((errors + 1)); }
-test -f agents/mnemosyne.agent.md || { echo "   ❌ mnemosyne.agent.md not found"; errors=$((errors + 1)); }
-test -f agents/zeus.agent.md || { echo "   ❌ zeus.agent.md not found"; errors=$((errors + 1)); }
-test -f routing.yml || { echo "   ❌ routing.yml not found"; errors=$((errors + 1)); }
-test -f skills/artifact-management/SKILL.md || { echo "   ❌ artifact-protocol.instructions.md not found"; errors=$((errors + 1)); }
-test -f .pantheon/memory-bank/01-active-context.md || { echo "   ❌ active-context.md not found"; errors=$((errors + 1)); }
-test -d .pantheon/memory-bank/.tmp || mkdir -p .pantheon/memory-bank/.tmp
-
-if [ -f scripts/scrub-secrets.py ]; then
-    echo "   ✅ scrub-secrets.py found (217 lines, $(grep -c 'def ' scripts/scrub-secrets.py || echo 0) functions)"
-else
-    echo "   ⚠️  scrub-secrets.py not found (will skip scrubbing)"
-fi
-
-# Verify compress_context handoff in routing.yml
-if grep -q "compress_context" routing.yml; then
-    HANDOFF_LINE=$(grep -n "compress_context" routing.yml | head -1 | cut -d: -f1)
-    echo "   ✅ compress_context handoff found in routing.yml at line $HANDOFF_LINE"
-else
-    echo "   ❌ compress_context handoff missing from routing.yml"
-    errors=$((errors + 1))
-fi
-
-# Verify ZZ artifact mention in artifact-protocol
-if grep -q "ZZ-phase" skills/artifact-management/SKILL.md; then
-    echo "   ✅ ZZ artifact format documented in artifact-protocol.instructions.md"
-else
-    echo "   ❌ ZZ artifact format missing from artifact-protocol.instructions.md"
-    errors=$((errors + 1))
-fi
-
-# Verify context-compression trigger was added to zeus.agent.md
-if grep -q "Context Compression Trigger" agents/zeus.agent.md; then
-    echo "   ✅ Context Compression Trigger section present in zeus.agent.md"
-else
-    echo "   ⚠️  Context Compression Trigger section not yet in zeus.agent.md"
-fi
-
-if [ "$errors" -gt 0 ]; then
-    echo "   ❌ $errors prerequisite(s) missing — aborting"
-    exit 1
-fi
-echo "   ✅ All prerequisites met"
-echo ""
-
-# Step 2: Create a mock IMPL artifact for testing
-echo "[2/4] Creating mock IMPL artifact..."
-
-cat > .pantheon/memory-bank/.tmp/IMPL-test-phase-hermes.md << 'ARTIFACT'
-# IMPL-test-phase-hermes
-**Date:** 2026-06-26  **Status:** Awaiting Compression Test
-
-## What Was Implemented
-- test/file.py — mock implementation for compression test
-- test/schema.py — mock schema migration for priority scoring test
-
-## Tests
-- ✅ 5 tests passing / Coverage: 87%
-
-## Notes for Themis
-This is a test artifact to validate the compression pipeline end-to-end.
-ARTIFACT
-
-cat > .pantheon/memory-bank/.tmp/IMPL-test-phase-demeter.md << 'ARTIFACT'
-# IMPL-test-phase-demeter
-**Date:** 2026-06-26  **Status:** Awaiting Compression Test
-
-## What Was Implemented
-- backend/migrations/0013_add_refresh_tokens.py — new refresh_tokens table
-- backend/models/token.py — Token model with FK to users
-
-## Tests
-- ✅ Migration tested: upgrade + downgrade both verified
-
-## Notes for Themis
-New table: refresh_tokens with (user_id, token_hash, expires_at, revoked_at)
-ARTIFACT
-
-echo "   ✅ 2 mock artifacts created (hermes + demeter)"
-echo ""
-
-# Step 3: Create mock subtask_summary for compression
-echo "[3/4] Creating mock subtask summaries..."
-
-cat > .pantheon/memory-bank/.tmp/subtask-hermes.json << 'JSON'
-{
-  "files_changed": ["test/file.py", "test/schema.py"],
-  "summary": "Added JWT auth endpoint with refresh token rotation",
-  "tests": "✅ All passing",
-  "coverage": "87%",
-  "status": "complete",
-  "blockers": null
-}
-JSON
-
-cat > .pantheon/memory-bank/.tmp/subtask-demeter.json << 'JSON'
-{
-  "files_changed": ["backend/migrations/0013_add_refresh_tokens.py", "backend/models/token.py"],
-  "summary": "Created refresh_tokens table with FK to users, hashed token, expires_at, revoked_at",
-  "tests": "✅ Migration upgrade + downgrade verified",
-  "coverage": "100%",
-  "status": "complete",
-  "blockers": null
-}
-JSON
-
-echo "   ✅ 2 mock subtask summaries created"
-echo ""
-
-# Step 4: Verify .tmp structure is writable and correct
-echo "[4/4] Verifying .tmp structure..."
-ls -la .pantheon/memory-bank/.tmp/
-echo ""
-echo "   ✅ .tmp structure looks correct"
-echo ""
-
-# Step 5: Validate scrub-secrets.py works
-echo "[5/4] Testing scrub-secrets.py (bonus)..."
-if [ -f scripts/scrub-secrets.py ]; then
-    # Test with a simple secret. The token is assembled at runtime so the
-    # shipped script never contains a literal `Bearer <token>` sequence (the
-    # redaction gate flags any such literal as a credential).
-    SCRUB_FIXTURE_TOKEN="my-test-token-12345"
-    SECRET_TEST=$(echo "Bearer ${SCRUB_FIXTURE_TOKEN}" | python3 scripts/scrub-secrets.py --stdin 2>/dev/null)
-    if echo "$SECRET_TEST" | grep -q "REDACTED"; then
-        echo "   ✅ scrub-secrets.py: secret detection working"
-    else
-        echo "   ⚠️  scrub-secrets.py: did not detect test secret (may need different pattern)"
-    fi
-fi
-echo ""
-
-echo "=== Test Complete ==="
-echo ""
-echo "Mock artifacts ready for compression pipeline test."
-echo ""
-echo "Next manual step: Trigger compress_context handoff:"
-echo "  @mnemosyne Run compression pipeline with test artifacts"
-echo ""
-echo "Expected output:"
-echo "  - .pantheon/memory-bank/.tmp/ZZ-test-context.md"
-echo "  - Updated 01-active-context.md with compressed entries"
-echo "  - Updated 02-progress-log.md with archived mock artifact"
-echo ""
-echo "Scoring verification (expected):"
-echo "  - subtask-hermes.json: JWT + endpoint + token keywords → HIGH (score ~0.71)"
-echo "  - subtask-demeter.json: migration + new table + FK keywords → CRITICAL (score ~0.83)"
-echo ""
-echo "To clean up test artifacts:"
-echo "  rm .pantheon/memory-bank/.tmp/IMPL-test-phase-*.md"
-echo "  rm .pantheon/memory-bank/.tmp/subtask-*.json"

package/src/pantheon/auto-wake.ts

@@ -1,85 +0,0 @@
-/**
- * Auto-Wake Consumer — scans signal files from the BackgroundJobBoard
- * and returns completed job signals for Zeus to process.
- *
- * Signal files are created by the board on every terminal state transition
- * (completed/error/cancelled). This module consumes them atomically by
- * reading the `.signal.json` content, then renaming to `.consumed.json`,
- * preventing race conditions where the same signal is processed twice.
- *
- * The rename is the commit point: if Zeus crashes between the read and the
- * rename, the `.signal.json` file remains and will be re-processed on restart.
- * This gives **at-least-once** delivery, which is acceptable because the board's
- * `markReconciled()` is idempotent.
- *
- * @module auto-wake
- */
-
-import { existsSync } from 'node:fs'
-import { readdir, readFile, rename } from 'node:fs/promises'
-import { join } from 'node:path'
-
-/** Signal structure written by BackgroundJobBoard.writeSignal(). */
-export interface WakeSignal {
-  taskID: string
-  alias: string
-  agent: string
-  state: string
-  summary: string | null
-  timestamp: number
-}
-
-/** Default signal directory (matches BoardOptions.signalDir default in plugin.ts). */
-const DEFAULT_SIGNAL_DIR = '.pantheon/deepwork/board-signals'
-
-/**
- * Scan the signal directory for `*.signal.json` files, atomically consume
- * each by renaming to `*.consumed.json`, and return the parsed signals.
- *
- * Safe to call multiple times — already-consumed files (`.consumed.json`)
- * are ignored, and the atomic rename prevents double-processing.
- *
- * The flow per signal file:
- * 1. Read the `.signal.json` content
- * 2. Rename `.signal.json` → `.consumed.json` (atomic: OS-level rename)
- * 3. If rename succeeds, push the signal into the result
- * 4. If rename fails (another process consumed it first), skip
- *
- * @param signalDir  Directory to scan (default: `.pantheon/deepwork/board-signals`)
- * @returns List of wake signals from newly consumed files
- */
-export async function consumeWakeSignals(signalDir?: string): Promise<WakeSignal[]> {
-  const dir = signalDir ?? DEFAULT_SIGNAL_DIR
-
-  // If the signal directory doesn't exist, there are no signals
-  if (!existsSync(dir)) {
-    return []
-  }
-
-  const entries = await readdir(dir, { withFileTypes: true })
-  const signals: WakeSignal[] = []
-
-  for (const entry of entries) {
-    if (!entry.isFile()) continue
-    if (!entry.name.endsWith('.signal.json')) continue
-
-    const signalPath = join(dir, entry.name)
-    const consumedName = entry.name.replace(/\.signal\.json$/, '.consumed.json')
-    const consumedPath = join(dir, consumedName)
-
-    try {
-      // Step 1: Read the file content
-      const content = await readFile(signalPath, 'utf-8')
-      const data = JSON.parse(content) as WakeSignal
-
-      // Step 2: Atomically rename — this is the commit point.
-      // If this fails, the .signal.json remains and we'll retry next call.
-      await rename(signalPath, consumedPath)
-
-      // Step 3: Only add to results after successful rename
-      signals.push(data)
-    } catch {}
-  }
-
-  return signals
-}

package/src/pantheon/chat-reminders.ts

@@ -1,9 +0,0 @@
-/**
- * Retired transcript-notification module.
- *
- * Delegation lifecycle state must not be queued for delivery into user
- * messages. Visibility is provided by the board/list/read tools and TUI
- * toasts; this intentionally exports no delivery API.
- *
- * @module chat-reminders
- */

package/src/pantheon/comment-checker.ts

@@ -1,48 +0,0 @@
-/**
- * Comment Checker — detects trivial/auto-generated comments that make code
- * look machine-produced rather than human-written.
- *
- * Used by @themis (Layer 1: Surface review) to flag comments that add no
- * information value and signal AI-generated code.
- */
-
-/** Regex patterns matching trivial, uninformative comments. */
-export const TRIVIAL_PATTERNS: readonly RegExp[] = [
-  /#\s*increment\s+\w+/i,
-  /#\s*decrement\s+\w+/i,
-  /#\s*set\s+\w+/i,
-  /#\s*loop\s+over/i,
-  /#\s*(initialize|init)\s+\w+/i,
-  /#\s*(return|yield)\s+\w+/i,
-]
-
-export interface CommentCheckResult {
-  /** Number of trivial comments found (0 = clean). */
-  score: number
-  /** Human-readable descriptions of each flagged comment. */
-  flags: string[]
-}
-
-/**
- * Scan source code for trivial comments that add no information value.
- *
- * Returns a score (count of trivial comments) and a list of flags describing
- * each match. A score of 0 means no trivial comments were detected.
- */
-export function checkCommentDensity(source: string): CommentCheckResult {
-  const flags: string[] = []
-  const lines = source.split('\n')
-
-  for (let i = 0; i < lines.length; i++) {
-    const line = lines[i] as string
-    for (const pattern of TRIVIAL_PATTERNS) {
-      if (pattern.test(line)) {
-        const trimmed = line.trim()
-        flags.push(`Line ${i + 1}: trivial comment — "${trimmed}"`)
-        break // one flag per line, even if multiple patterns match
-      }
-    }
-  }
-
-  return { score: flags.length, flags }
-}

package/src/pantheon/compaction-assert.ts

@@ -1,160 +0,0 @@
-/**
- * Compaction Re-Assert (release-134 Phase 4) — detect fresh state after
- * compaction without writing to the chat transcript.
- *
- * WHY: after a session is compacted, the system state may have changed since
- * the summary was built (new background delegations launched, results
- * delivered, goals updated). The compacted session's context is frozen at
- * summary time — the model would operate on stale state.
- *
- * HOW: the compaction lifecycle keeps the board and goal stores authoritative;
- * the `experimental.session.compacting` hook builds the context that is
- * consumed by the compaction itself. Chat is never used as a delivery channel.
- *
- * Deliberately NOT `session.promptAsync({noReply:true})`: the sprint 4+7
- * spike refuted noReply as a push mechanism on this opencode build (nothing
- * is delivered to the parent).
- *
- * FAIL-OPEN: every source (board list and goals list) is wrapped —
- * a failure logs a warn via the silence-by-default logger and the hook never
- * throws. A session with NO fresh state (nothing running/unread/goals) is a
- * SILENT skip (no warn). The block is capped at
- * REASSERT_MAX_LINES item lines (1 header) so it stays compact.
- *
- * Pure TypeScript — zero runtime deps beyond the board/goal types.
- *
- * @module compaction-assert
- */
-import type { BackgroundJobBoard } from './background-job-board.ts'
-import type { CompactionGoalSource } from './delegation-compaction.ts'
-import { createPantheonLogger } from './logger.ts'
-
-// Silence-by-default TUI policy (pantheon-hooks L42-58): console echo is
-// opt-in via PANTHEON_HOOKS_LOG=1. `deps.logger` injection stays for tests.
-const log = createPantheonLogger({ module: 'pantheon-compaction' })
-
-// ─── Constants ─────────────────────────────────────────────────────────
-
-/** Max fresh-state item lines (running + unread + goals) in the block. */
-export const REASSERT_MAX_LINES = 10
-
-/** Header line — tells the model the summary may be stale. */
-export const REASSERT_HEADER = 'State re-assertion after compaction — the summary may be stale:'
-
-// ─── Types ─────────────────────────────────────────────────────────────
-
-/** Dependencies threaded to reassertAfterCompaction (wired from plugin.ts). */
-export interface CompactionAssertDeps {
-  /** The session whose compaction context is being prepared. */
-  sessionID: string
-  /** Board query surface — `list` is all the re-assertion needs. */
-  board: Pick<BackgroundJobBoard, 'list'>
-  /** Goal source for the goals block (optional — same gating as Phase 2). */
-  goals?: CompactionGoalSource
-  /** Injectable warn logger (testable). Defaults to silence-by-default. */
-  logger?: { warn: (message: string) => void }
-  /** Non-chat delivery sink (for example a pending `output.context` block). */
-  deliverContext?: (context: string[]) => void | Promise<void>
-}
-
-// ─── Formatting ────────────────────────────────────────────────────────
-
-/** One-line terminal status label, mirroring formatForPrompt. */
-function terminalLabel(state: string): string {
-  switch (state) {
-    case 'completed':
-      return 'OK'
-    case 'error':
-      return 'ERR'
-    case 'cancelled':
-      return 'CAN'
-    default:
-      return state.toUpperCase()
-  }
-}
-
-/** Collapse whitespace and truncate a description/objective. */
-function truncate(text: string, max = 80): string {
-  const flat = text.trim().replace(/\s+/g, ' ')
-  return flat.length <= max ? flat : `${flat.slice(0, max)}…`
-}
-
-// ─── Build ─────────────────────────────────────────────────────────────
-
-/**
- * Build the compact fresh-state lines: running delegations first (active work
- * is the freshest signal), then unread terminal delegations, then active
- * goals (only while the line budget allows). Returns undefined when there is
- * nothing to assert (silent skip) or a source failed (warn + skip).
- */
-async function buildStateLines(
-  deps: CompactionAssertDeps,
-  warn: (message: string) => void,
-): Promise<string[] | undefined> {
-  const lines: string[] = []
-  try {
-    const jobs = deps.board.list(deps.sessionID)
-    const running = jobs
-      .filter((j) => j.state === 'running')
-      .sort((a, b) => a.launchedAt - b.launchedAt)
-    const unread = jobs
-      .filter((j) => j.terminalUnreconciled)
-      .sort((a, b) => b.updatedAt - a.updatedAt)
-
-    for (const job of running) {
-      if (lines.length >= REASSERT_MAX_LINES) break
-      lines.push(`  running [${job.alias}] ${job.agent} — ${truncate(job.description)}`)
-    }
-    for (const job of unread) {
-      if (lines.length >= REASSERT_MAX_LINES) break
-      lines.push(
-        `  unread [${job.alias}] ${job.agent} — ${truncate(job.description)} — ${terminalLabel(job.state)}`,
-      )
-    }
-
-    const source = deps.goals
-    if (lines.length < REASSERT_MAX_LINES && source !== undefined && source.enabled) {
-      const goals = await source.list(deps.sessionID)
-      for (const goal of goals.filter((g) => g.status !== 'done')) {
-        if (lines.length >= REASSERT_MAX_LINES) break
-        lines.push(`  goal [${goal.id}] ${truncate(goal.objective)} — ${goal.status}`)
-      }
-    }
-  } catch (err: unknown) {
-    const reason = err instanceof Error ? err.message : String(err)
-    warn(`compaction re-assert failed for session ${deps.sessionID}: ${reason}`)
-    return undefined
-  }
-  return lines.length === 0 ? undefined : lines
-}
-
-/**
- * Re-assert fresh state for a compacted session and return the context block
- * to the non-chat caller. The caller can hand this array to OpenCode's
- * `output.context` or persist it through the board; constructing it here and
- * throwing it away would make the hook a no-op.
- *
- * Returns the fresh-state context when found, or undefined when there was
- * nothing to assert (silent skip) or a source failed (warned). NEVER throws —
- * the event hook must never break the session.
- */
-export async function reassertAfterCompaction(
-  deps: CompactionAssertDeps,
-): Promise<string[] | undefined> {
-  // Silence-by-default logger: file-only warn, console echo opt-in via
-  // PANTHEON_HOOKS_LOG=1 (TUI pollution policy, pantheon-hooks L42-58).
-  const warn = deps.logger?.warn ?? ((message: string) => log.warn(message))
-  try {
-    const lines = await buildStateLines(deps, warn)
-    if (lines === undefined) return undefined
-    // This is a context return value, not a chat message. The caller owns the
-    // supported delivery path (output.context or persisted board state).
-    const context = [REASSERT_HEADER, ...lines]
-    await deps.deliverContext?.(context)
-    return context
-  } catch (err: unknown) {
-    const reason = err instanceof Error ? err.message : String(err)
-    warn(`compaction re-assert failed for session ${deps.sessionID}: ${reason}`)
-    return undefined
-  }
-}