opencodekit 0.18.6 → 0.18.8

package/dist/index.js

@@ -18,7 +18,7 @@ var __require = /* @__PURE__ */ createRequire(import.meta.url);
 
 //#endregion
 //#region package.json
-var version = "0.18.6";
+var version = "0.18.8";
 
 //#endregion
 //#region src/utils/errors.ts

package/dist/template/.opencode/AGENTS.md

@@ -100,6 +100,14 @@ If a lookup, search, or tool call returns empty, partial, or suspiciously narrow
 - For lists, batches, or paginated results: determine expected scope, track processed items, confirm full coverage
 - If any item is blocked by missing data, mark it `[blocked]` and state exactly what is missing
 
+### Plan Quality Gate
+
+Before approving or executing any implementation plan:
+
+1. Plan MUST contain a `## Discovery` section with substantive research findings (>100 characters)
+2. Plans without documented discovery skip the research phase and produce worse implementations
+3. If discovery is missing or boilerplate, reject the plan and research first
+
 ---
 
 ## Hard Constraints (Never Violate)
@@ -157,6 +165,74 @@ Use specialist agents by intent:
 
 **Parallelism rule**: Use parallel subagents for 3+ independent tasks; otherwise work sequentially.
 
+### Worker Distrust Protocol
+
+Subagent self-reports are **approximately 50% accurate**. After every `task()` returns:
+
+1. **Read changed files directly** — don't trust the summary; `git diff` or read modified files
+2. **Run verification on modified files** — typecheck + lint at minimum; tests if the change touches behavior
+3. **Check acceptance criteria** — compare actual output against the original task spec, not the agent's claims
+4. **Verify nothing was broken** — check that files outside the agent's scope weren't unexpectedly modified
+
+```
+✅ Agent reports success → Read diff → Run verification → Confirm criteria → Accept
+❌ Agent reports success → Trust it → Move on
+❌ Agent reports success → Skim summary → Accept
+```
+
+This applies to ALL subagent types (`@general`, `@explore`, `@review`, `@scout`), not just implementation agents.
+
+### Structured Termination Contract
+
+Every subagent task MUST return a structured response. When dispatching, include this in the prompt:
+
+```
+Return your results in this exact format:
+
+## Result
+- **Status:** completed | blocked | failed
+- **Files Modified:** [list of file paths]
+- **Files Read:** [list of file paths consulted]
+
+## Verification
+- [What you verified and how]
+- [Command output or evidence]
+
+## Summary
+[2-5 sentences: what was done, key decisions, anything unexpected]
+
+## Blockers (if status is blocked/failed)
+- [What's blocking]
+- [What was tried]
+- [Recommended next step]
+```
+
+When a subagent returns WITHOUT this structure, treat the response with extra skepticism — unstructured reports are more likely to omit failures or exaggerate completion.
+
+### Context File Pattern
+
+For complex delegations, write context to a file instead of inlining it in the `task()` prompt:
+
+```typescript
+// ❌ Token-expensive: inlining large context
+task({
+  prompt: `Here is the full plan:\n${longPlanContent}\n\nImplement task 3...`
+});
+
+// ✅ Token-efficient: reference by path
+// Write context file first:
+write('.beads/artifacts/<id>/worker-context.md', contextContent);
+// Then reference it:
+task({
+  prompt: `Read the context file at .beads/artifacts/<id>/worker-context.md\n\nImplement task 3 as described in that file.`
+});
+```
+
+Use this pattern when:
+- Context exceeds ~500 tokens
+- Multiple subagents need the same context
+- Plan content, research findings, or specs need to be passed to workers
+
 ---
 
 ## Question Policy
@@ -201,19 +277,21 @@ For major tracked work:
 | Phase             | Target  | Action                                     |
 | ----------------- | ------- | ------------------------------------------ |
 | Starting work     | <50k    | Load only essential AGENTS.md + task spec  |
-| Mid-task          | 50-100k | Distill completed reads, keep active files |
-| Approaching limit | >100k   | Aggressive distill, prune remaining noise  |
+| Mid-task          | 50-100k | Compress completed phases, keep active files |
+| Approaching limit | >100k   | Aggressive compression, sweep stale noise    |
 | Near capacity     | >150k   | Session restart with handoff               |
 
-### Tools
+### DCP Commands
 
-- `distill` — Extract key info from tool outputs, then remove raw output (preferred)
-- `prune` — Remove tool outputs entirely (noise only, no preservation)
+- `/dcp context` — Show current context health and pressure
+- `/dcp compress` — Compress completed conversation ranges (primary tool)
+- `/dcp sweep` — Remove stale/noisy content according to DCP rules
+- `/dcp stats` — Inspect pruning/compression activity
 
 ### Rules
 
-1. **Distill at turn START** — not end (you know what's needed)
-2. **Batch operations** — accumulate candidates before acting
+1. **Compress at phase boundaries** — not during active edits
+2. **Batch cleanup** — use `/dcp sweep` for stale noise, not ad-hoc deletion
 3. **Protected content** — AGENTS.md, .opencode/, .beads/, config files
 
 ---

package/dist/template/.opencode/agent/build.md

@@ -350,6 +350,15 @@ task({ subagent_type: "general", description: "...", prompt: "..." });
 
 Then synthesize results, verify locally, and report with file-level evidence.
 
+**Mandatory post-delegation steps** (Worker Distrust Protocol):
+
+1. Read changed files directly — don't trust the agent summary
+2. Run verification on modified files (typecheck + lint minimum)
+3. Check acceptance criteria against the original task spec
+4. Only then accept the work
+
+Include the **Structured Termination Contract** in every subagent prompt (Result/Verification/Summary/Blockers format). See AGENTS.md delegation policy for the template.
+
 ## Output
 
 Report in this order:

package/dist/template/.opencode/dcp.jsonc

@@ -1,83 +1,72 @@
 {
-  "$schema": "https://raw.githubusercontent.com/Opencode-DCP/opencode-dynamic-context-pruning/master/dcp.schema.json",
-  "enabled": true,
-  "debug": false,
-  // "minimal" shows prune activity without noise; "detailed" shows token counts
-  "pruneNotification": "detailed",
-  // "chat" (in-conversation) or "toast" (system notification)
-  "pruneNotificationType": "toast",
-  // Commands: /dcp context, /dcp stats, /dcp sweep
-  "commands": {
-    "enabled": true,
-    // Protect these from /dcp sweep
-    "protectedTools": ["observation", "memory-update", "memory-search"],
-  },
-  "turnProtection": {
-    "enabled": true,
-    "turns": 4,
-  },
-  // Protected file patterns - never auto-prune reads of these files
-  "protectedFilePatterns": [
-    "**/.env*",
-    "**/AGENTS.md",
-    "**/.opencode/**",
-    "**/.beads/**",
-    "**/package.json",
-    "**/tsconfig.json",
-    "**/biome.json",
-  ],
-  "tools": {
-    "settings": {
-      "nudgeEnabled": true,
-      "nudgeFrequency": 10,
-      // v2.0.1: percentage support - 80% of model's context (most models = 128k)
-      "contextLimit": "80%",
-      // Protect state-modifying and critical workflow tools
-      // LSP excluded: ephemeral exploration, prune after understanding
-      "protectedTools": [
-        "write",
-        "edit",
-        "memory-update",
-        "observation",
-        "skill",
-        "skill_mcp",
-        "task",
-        "batch",
-        "todowrite",
-        "todoread",
-      ],
-    },
-    // v2.0.0: permission model - "allow", "ask", or "deny"
-    "prune": {
-      "permission": "allow",
-    },
-    "distill": {
-      "permission": "allow",
-      "showDistillation": false,
-    },
-    "compress": {
-      // v2.1.1: compress can have boundary matching issues - use ask for safety
-      "permission": "allow",
-      "showCompression": false,
-    },
-  },
-  // Auto strategies - TOP LEVEL, not under tools
-  "strategies": {
-    // Dedup = zero LLM cost, high impact - always enable
-    "deduplication": {
-      "enabled": true,
-      "protectedTools": [],
-    },
-    // Supersede writes = zero cost, removes redundant write inputs after read
-    // Note: default changed to false in beta, we explicitly enable
-    "supersedeWrites": {
-      "enabled": true,
-    },
-    // Purge error inputs after N turns
-    "purgeErrors": {
-      "enabled": true,
-      "turns": 4,
-      "protectedTools": [],
-    },
-  },
+	"$schema": "https://raw.githubusercontent.com/Opencode-DCP/opencode-dynamic-context-pruning/master/dcp.schema.json",
+	"enabled": true,
+	"debug": false,
+	// "minimal" shows prune activity without noise; "detailed" shows token counts
+	"pruneNotification": "detailed",
+	// "chat" (in-conversation) or "toast" (system notification)
+	"pruneNotificationType": "toast",
+	// Commands: /dcp context, /dcp stats, /dcp sweep, /dcp compress, /dcp decompress, /dcp recompress
+	"commands": {
+		"enabled": true,
+		// Protect these from /dcp sweep
+		"protectedTools": ["observation", "memory-update", "memory-search"]
+	},
+	// Manual mode: disables autonomous context management
+	"manualMode": {
+		"enabled": false,
+		"automaticStrategies": true
+	},
+	"turnProtection": {
+		"enabled": true,
+		"turns": 4
+	},
+	// Protected file patterns - never auto-prune reads of these files
+	"protectedFilePatterns": [
+		"**/.env*",
+		"**/AGENTS.md",
+		"**/.opencode/**",
+		"**/.beads/**",
+		"**/package.json",
+		"**/tsconfig.json",
+		"**/biome.json"
+	],
+	"compress": {
+		// v3.0.0: single compress tool replaces the old 3-tool system
+		"permission": "allow",
+		"showCompression": false,
+		"maxContextLimit": "80%",
+		"minContextLimit": 30000,
+		"nudgeFrequency": 5,
+		"iterationNudgeThreshold": 15,
+		"nudgeForce": "soft",
+		"flatSchema": false,
+		"protectUserMessages": true,
+		// v3.0.0 auto-protects: task, skill, todowrite, todoread, compress, batch, plan_enter, plan_exit
+		// Only list additional tools here
+		"protectedTools": ["write", "edit", "memory-*", "observation", "tilth_*"]
+	},
+	// Experimental features
+	"experimental": {
+		"allowSubAgents": true,
+		"customPrompts": false
+	},
+	// Auto strategies
+	"strategies": {
+		// Dedup = zero LLM cost, high impact - always enable
+		"deduplication": {
+			"enabled": true,
+			"protectedTools": []
+		},
+		// Supersede writes = zero cost, removes redundant write inputs after read
+		"supersedeWrites": {
+			"enabled": true
+		},
+		// Purge error inputs after N turns
+		"purgeErrors": {
+			"enabled": true,
+			"turns": 4,
+			"protectedTools": []
+		}
+	}
 }

package/dist/template/.opencode/memory/project/gotchas.md

@@ -38,10 +38,10 @@ The edit tool (`str_replace`) is the #1 source of failures in LLM coding. Models
 
 ### Context Hygiene
 
-- Distill large tool outputs immediately after use
-- Prune irrelevant reads before proceeding
+- Compress completed work phases before moving on
+- Use `/dcp sweep` after a closed phase to remove stale noise
 - Token budget: <50k start → 50-100k mid → >150k restart session
-- Subagent outputs can leak tokens — be aggressive about distilling
+- Subagent outputs can leak tokens — compress completed phases and sweep stale subagent noise
 
 ## OpenCode Config
 

package/dist/template/.opencode/package.json

@@ -11,7 +11,7 @@
     "type-check": "tsc --noEmit"
   },
   "dependencies": {
-    "@opencode-ai/plugin": "1.2.21"
+    "@opencode-ai/plugin": "1.2.24"
   },
   "devDependencies": {
     "@types/node": "^25.3.0",

package/dist/template/.opencode/pnpm-lock.yaml

@@ -0,0 +1,71 @@
+lockfileVersion: '9.0'
+
+settings:
+  autoInstallPeers: true
+  excludeLinksFromLockfile: false
+
+importers:
+
+  .:
+    dependencies:
+      '@opencode-ai/plugin':
+        specifier: 1.2.22
+        version: 1.2.22
+    devDependencies:
+      '@types/node':
+        specifier: ^25.3.0
+        version: 25.3.5
+      bun-types:
+        specifier: ^1.3.9
+        version: 1.3.10
+      typescript:
+        specifier: ^5.9.3
+        version: 5.9.3
+
+packages:
+
+  '@opencode-ai/plugin@1.2.22':
+    resolution: {integrity: sha512-vKVHymAGspLDBuObiJNSoCAQR+l6aMx6n21mgwNBa0BY4n0uzdJcM1rAN/Ksv/t6rsswMcsNeH0yMmj9Iokf7Q==}
+
+  '@opencode-ai/sdk@1.2.22':
+    resolution: {integrity: sha512-EXI1+Rc9dkzKo9uDj0EA4pLUCQJS2Qi8/xKwsT6RNZPBpSau05HVPWlZ+t8WyLW4fWymiho8jsXSErfKOpyCsw==}
+
+  '@types/node@25.3.5':
+    resolution: {integrity: sha512-oX8xrhvpiyRCQkG1MFchB09f+cXftgIXb3a7UUa4Y3wpmZPw5tyZGTLWhlESOLq1Rq6oDlc8npVU2/9xiCuXMA==}
+
+  bun-types@1.3.10:
+    resolution: {integrity: sha512-tcpfCCl6XWo6nCVnpcVrxQ+9AYN1iqMIzgrSKYMB/fjLtV2eyAVEg7AxQJuCq/26R6HpKWykQXuSOq/21RYcbg==}
+
+  typescript@5.9.3:
+    resolution: {integrity: sha512-jl1vZzPDinLr9eUt3J/t7V6FgNEw9QjvBPdysz9KfQDD41fQrC2Y4vKQdiaUpFT4bXlb1RHhLpp8wtm6M5TgSw==}
+    engines: {node: '>=14.17'}
+    hasBin: true
+
+  undici-types@7.18.2:
+    resolution: {integrity: sha512-AsuCzffGHJybSaRrmr5eHr81mwJU3kjw6M+uprWvCXiNeN9SOGwQ3Jn8jb8m3Z6izVgknn1R0FTCEAP2QrLY/w==}
+
+  zod@4.1.8:
+    resolution: {integrity: sha512-5R1P+WwQqmmMIEACyzSvo4JXHY5WiAFHRMg+zBZKgKS+Q1viRa0C1hmUKtHltoIFKtIdki3pRxkmpP74jnNYHQ==}
+
+snapshots:
+
+  '@opencode-ai/plugin@1.2.22':
+    dependencies:
+      '@opencode-ai/sdk': 1.2.22
+      zod: 4.1.8
+
+  '@opencode-ai/sdk@1.2.22': {}
+
+  '@types/node@25.3.5':
+    dependencies:
+      undici-types: 7.18.2
+
+  bun-types@1.3.10:
+    dependencies:
+      '@types/node': 25.3.5
+
+  typescript@5.9.3: {}
+
+  undici-types@7.18.2: {}
+
+  zod@4.1.8: {}

package/dist/template/.opencode/skill/compaction/SKILL.md

@@ -13,7 +13,7 @@ dependencies: []
 
 ## When to Use
 
-- Context is growing large and needs distillation/compaction to continue.
+- Context is growing large and needs compression/cleanup to continue.
 
 ## When NOT to Use
 
@@ -33,8 +33,8 @@ Long-running sessions accumulate context (tool outputs, code reads, exploration
 | Context Usage | Status      | Action                                        |
 | ------------- | ----------- | --------------------------------------------- |
 | 0-50%         | 🟢 Normal   | Work freely                                   |
-| 50-70%        | 🟡 Watch    | Start distilling completed explorations       |
-| 70-85%        | 🟠 Compact  | Actively compress/prune, consider handoff     |
+| 50-70%        | 🟡 Watch    | Start compressing completed explorations      |
+| 70-85%        | 🟠 Compact  | Actively compress and sweep stale noise       |
 | 85-95%        | 🔴 Critical | Emergency compaction, prepare session handoff |
 | 95%+          | ⛔ Limit    | Session handoff required                      |
 
@@ -42,31 +42,17 @@ Long-running sessions accumulate context (tool outputs, code reads, exploration
 
 Pay attention to these signals:
 
-- Tool outputs accumulating without being distilled
+- Completed phases accumulating without being compressed
 - Repeated file reads of the same content
 - Large bash outputs from builds/tests
 - Multiple exploration rounds without synthesis
 
 ## Compaction Strategies
 
-### Strategy 1: Proactive Distillation (Preferred)
+### Strategy 1: Phase Compression (Preferred)
 
-Distill tool outputs as you finish using them. This is the most granular and least lossy approach.
+Compress completed conversation phases into dense summaries. This is the primary DCP instrument in the installed beta.
 
-```
-WHEN: You've read a file and extracted what you need
-DO: distill({ targets: [{ id: "X", distillation: "..." }] })
-
-WHEN: Bash output gave you the answer you needed
-DO: distill({ targets: [{ id: "Y", distillation: "..." }] })
-
-WHEN: Search results identified the relevant files
-DO: distill({ targets: [{ id: "Z", distillation: "..." }] })
-```
-
-**Key principle**: Distill when you're DONE with the raw output, not while you still need it.
-
-### Strategy 2: Phase Compression
 
 Compress completed conversation phases into dense summaries.
 
@@ -84,21 +70,18 @@ DO: compress({
 
 **Key principle**: Only compress CLOSED chapters. Never compress active work.
 
-### Strategy 3: Noise Pruning
+### Strategy 2: Sweep Stale Noise
 
-Remove tool outputs that add zero value.
+Use `/dcp sweep` after a phase is complete to remove stale/noisy content automatically.
 
 ```
-WHEN: Tool output was irrelevant (wrong file, empty search results)
-DO: prune({ ids: ["X", "Y", "Z"] })
-
-WHEN: Earlier output is superseded by newer data
-DO: prune({ ids: ["old_id"] })
+WHEN: Wrong-target searches or superseded outputs are no longer needed
+DO: /dcp sweep
 ```
 
-**Key principle**: Prune noise, not signal. If in doubt, keep it.
+**Key principle**: Sweep only after the relevant phase is closed. If in doubt, keep it.
 
-### Strategy 4: Session Handoff
+### Strategy 3: Session Handoff
 
 When context is too large to compact further, hand off to a new session.
 
@@ -117,9 +100,8 @@ DO:
 Is context growing large?
 ├── NO → Continue working normally
 └── YES → What type of content is consuming space?
-    ├── Tool outputs I'm done with → DISTILL
     ├── Completed conversation phases → COMPRESS
-    ├── Irrelevant/superseded outputs → PRUNE
+    ├── Stale/noisy closed-phase outputs → SWEEP
     └── Everything is still relevant → SESSION HANDOFF
 ```
 
@@ -275,9 +257,8 @@ memory - search({ query: "auth implementation" });
 
 This project uses `@tarquinen/opencode-dcp` for always-on context management (injected via `experimental.chat.system.transform`):
 
-- **distill**: High-fidelity extraction from tool outputs (favored instrument)
-- **compress**: Phase-level conversation compression (sledgehammer — completed phases only)
-- **prune**: Targeted removal of noise (last resort — batch wisely)
+- **compress**: Phase-level conversation compression (primary DCP tool in the installed beta)
+- **/dcp sweep**: Cleanup command for stale/noisy content after a phase is complete
 - **Prunable-tools list**: Auto-injected into messages with token estimates
 - **Nudge system**: Reminders every N tool calls + critical limit warnings
 
@@ -298,15 +279,6 @@ edit({ ... })      // Now you can't find the right location
 
 **Fix**: Keep raw content while actively editing. Compress AFTER the edit phase.
 
-### ❌ Lossy Distillation
-
-```
-// DON'T distill without capturing key details
-distill({ distillation: "Read the auth file, it has some functions" })
-```
-
-**Fix**: Include function signatures, types, key logic, constraints — everything you'd need to avoid re-reading.
-
 ### ❌ Compressing Active Work
 
 ```
@@ -324,23 +296,22 @@ compress({ summary: "Explored auth options" })
 // By the time you notice, emergency compaction loses information
 ```
 
-**Fix**: Monitor regularly. Distill as you go. Compress at natural breakpoints.
+**Fix**: Monitor regularly. Compress at natural breakpoints and use `/dcp sweep` after closed phases.
 
 ## Checklist
 
 Before compacting:
 
 - [ ] Identified what type of content is consuming context
-- [ ] Chosen appropriate strategy (distill/compress/prune/handoff)
+- [ ] Chosen appropriate strategy (compress/sweep/handoff)
 - [ ] Verified raw content is no longer needed for active work
-- [ ] Captured all key details in distillation/summary
+- [ ] Captured all key details in the compression summary or handoff
 - [ ] Saved important decisions as observations
 - [ ] Created handoff document if switching sessions
 
 During long sessions:
 
-- [ ] Distilling tool outputs after extracting insights
 - [ ] Compressing completed phases at natural breakpoints
-- [ ] Pruning noise and superseded outputs
+- [ ] Sweeping stale/noisy outputs after phase completion
 - [ ] Monitoring context usage trends
 - [ ] Planning session handoff if approaching limits

package/dist/template/.opencode/skill/context-engineering/SKILL.md

@@ -62,7 +62,7 @@ Extend it by:
 | Reading entire files    | Use `lsp documentSymbol` for outline    |
 | Loading whole documents | Read specific line ranges               |
 | Flat file loading       | Navigate AGENTS.md hierarchy            |
-| Keeping completed work  | Prune aggressively (context-management) |
+| Keeping completed work  | Compress closed phases, sweep stale noise (context-management) |
 
 ## Anti-Patterns
 

package/dist/template/.opencode/skill/context-management/SKILL.md

@@ -29,14 +29,14 @@ Use this skill to keep context useful from first turn to final handoff.
 Prefer **phase-level compression** over reactive cleanup.
 
 ```text
-compress > distill > prune
+compress > sweep > handoff
 ```
 
 - **compress**: Best default when a phase is complete
-- **distill**: Use when you must preserve technical detail but can remove bulky raw outputs
-- **prune**: Use only for true noise you are certain will never be needed
+- **sweep**: Use `/dcp sweep` to clean stale/noisy content after completion
+- **handoff**: Use when context pressure remains high even after compression
 
-## DCP Tool Usage
+## DCP Command Usage
 
 ### `/dcp compress`
 
@@ -46,26 +46,17 @@ Use for completed chapters of work (research, implementation wave, review sweep)
 - Lowest cognitive overhead on later turns
 - Usually lowest risk of deleting needed details
 
-### `/dcp distill`
+### `/dcp sweep`
 
-Use when raw output is large but details still matter later.
+Use after a phase is complete to let DCP remove stale/noisy content automatically.
 
-Include concrete facts in distillation:
-
-- function signatures
-- constraints and assumptions
-- file paths and key decisions
-- verification outcomes
-
-### `/dcp prune`
-
-Use only for irrelevant/noise outputs:
+Good candidates for sweep-driven cleanup:
 
 - wrong-target searches
 - failed dead-end exploration no longer needed
 - duplicate or superseded junk output
 
-Do **not** prune because output is "long". Length alone is not noise.
+Do **not** sweep because output is merely "long". Length alone is not noise.
 
 ## Session Lifecycle Protocol
 
@@ -82,7 +73,7 @@ Do **not** prune because output is "long". Length alone is not noise.
 
 - Keep active files readable until edits are done
 - At each natural boundary, evaluate compress candidates
-- Distill high-value technical outputs before removal
+- Compress completed phases before starting the next chapter
 
 ### 3) Pre-Handoff / Closeout
 
@@ -104,7 +95,7 @@ Use these thresholds as operational triggers:
 | --- | --- | --- |
 | <50k | Healthy start | Keep inputs minimal, avoid unnecessary reads |
 | 50k–100k | Moderate growth | Compress completed phases, keep active files intact |
-| >100k | High pressure | Aggressively compress by phase; distill critical leftovers |
+| >100k | High pressure | Aggressively compress by phase; run `/dcp sweep` on stale noise |
 | >150k | Near capacity | Perform handoff and resume in a fresh session |
 
 Secondary guardrails:
@@ -140,7 +131,7 @@ Carry forward decisions and constraints, not every intermediate log.
 | Anti-Pattern | Why It Hurts | Correct Pattern |
 | --- | --- | --- |
 | Compressing active work areas (losing precision needed for edits) | Removes exact lines needed for safe edits | Keep active file/tool outputs until edit + verification complete |
-| Pruning tool outputs you'll need later | Forces rework and increases error risk | Distill first, then remove raw output |
+| Sweeping content you still need | Forces rework and increases error risk | Keep active files/tool outputs until phase is complete |
 | Not compressing completed exploration phases | Bloats context and degrades later turns | Compress immediately at phase completion |
 | Session handoff without persisting key decisions to memory | Next session loses rationale and constraints | Write observations/memory updates before handoff |
 
@@ -151,7 +142,7 @@ Check context health: are completed phases compressed? Are active files still re
 Before claiming cleanup done, confirm:
 
 - Active edit targets are still present in readable form
-- Completed phases are compressed/distilled
+- Completed phases are compressed or intentionally kept for active work
 - No critical decision exists only in transient tool output
 - Handoff includes next actions and blockers
 
@@ -160,11 +151,10 @@ Before claiming cleanup done, confirm:
 ```text
 1) Start turn: /dcp context
 2) Identify completed phase ranges
-3) compress completed ranges
-4) distill high-value technical outputs
-5) prune true noise only
-6) persist key decisions to memory
-7) handoff/resume with focused rehydration
+3) /dcp compress completed ranges
+4) /dcp sweep stale/noisy outputs after phase completion
+5) persist key decisions to memory
+6) handoff/resume with focused rehydration
 ```
 
 ## See Also

package/dist/template/.opencode/skill/dispatching-parallel-agents/SKILL.md

@@ -82,6 +82,7 @@ Good agent prompts are:
 1. **Focused** - One clear problem domain
 2. **Self-contained** - All context needed to understand the problem
 3. **Specific about output** - What should the agent return?
+4. **Structured termination** - Include the Structured Termination Contract from AGENTS.md (Result/Verification/Summary/Blockers format)
 
 ```markdown
 Fix the 3 failing tests in src/agents/agent-tool-abort.test.ts:
@@ -101,9 +102,11 @@ These are timing/race condition issues. Your task:
 
 Do NOT just increase timeouts - find the real issue.
 
-Return: Summary of what you found and what you fixed.
+[Include Structured Termination Contract here]
 ```
 
+For large investigations (context >500 tokens), use the **Context File Pattern** from AGENTS.md — write context to `.beads/artifacts/<id>/investigation-context.md` and reference by path in the dispatch prompt.
+
 ## Common Mistakes
 
 **❌ Too broad:** "Fix all the tests" - agent gets lost

package/dist/template/.opencode/skill/requesting-code-review/SKILL.md

@@ -94,6 +94,36 @@ Choose the smallest review depth that still covers the risk.
 
 Do **not** dispatch 5 reviewers by reflex for every tiny change. That produces noise, not rigor.
 
+## Review Scope Self-Check
+
+Before writing ANY review finding, the reviewer MUST ask:
+
+> **"Am I questioning APPROACH or DOCUMENTATION/CORRECTNESS?"**
+
+| Answer | Action |
+| --- | --- |
+| **APPROACH** — "I would have done it differently" | **Stay silent.** The approach was decided during planning. Review doesn't re-litigate design. |
+| **DOCUMENTATION** — "A new developer couldn't execute this" | **Raise it.** Missing file paths, vague acceptance criteria, assumed context. |
+| **CORRECTNESS** — "This will break at runtime" | **Raise it.** Bugs, security holes, type errors, logic flaws. |
+
+### Red Flags for Scope Creep
+
+- Suggesting alternative libraries or frameworks → **out of scope** (approach)
+- "I would rename this to..." without a concrete bug → **out of scope** (style preference)
+- "Consider adding feature X" that wasn't in requirements → **out of scope** (scope inflation)
+- "This could be more performant" without evidence of a real problem → **out of scope** (premature optimization)
+
+### What Reviewers SHOULD Flag
+
+- Missing error handling that will crash in production → **correctness**
+- Vague acceptance criteria that workers can't verify → **documentation**
+- Hardcoded values with no explanation → **documentation**
+- Missing file paths for tasks that require edits → **documentation**
+- "Use the existing pattern" without specifying which pattern → **documentation**
+- Security vulnerabilities (injection, auth bypass, secrets) → **correctness**
+
+Include this self-check instruction in EVERY review dispatch prompt to prevent bikeshedding.
+
 ## Step 1: Get Git Context
 
 ### Setup Checklist

package/dist/template/.opencode/skill/subagent-driven-development/SKILL.md

@@ -35,6 +35,8 @@ dependencies: [executing-plans]
 
 Read plan file, create TodoWrite with all tasks.
 
+**Context file pattern:** If the plan exceeds ~500 tokens, write it to `.beads/artifacts/<id>/plan-context.md` and reference by path in subagent prompts instead of inlining. This saves tokens when dispatching multiple subagents from the same plan.
+
 ### 2. Execute Task with Subagent
 
 For each task:
@@ -56,10 +58,15 @@ Task tool (general-purpose):
 
     Work from: [directory]
 
-    Report: What you implemented, what you tested, test results, files changed, any issues
+    [Include Structured Termination Contract from AGENTS.md]
 ```
 
-**Subagent reports back** with summary of work.
+**After subagent reports back** — follow the **Worker Distrust Protocol** from AGENTS.md:
+
+1. Read changed files directly (don't trust the report)
+2. Run verification on modified files (typecheck + lint minimum)
+3. Check acceptance criteria against original task spec
+4. Only then mark the task as complete
 
 ### 3. Review Subagent's Work
 

package/dist/template/.opencode/skill/verification-before-completion/SKILL.md

@@ -168,6 +168,37 @@ If you just verified and nothing changed, don't re-verify:
 
 This matters when other commands need verification (e.g., closing beads, `/ship`). If you verified 30 seconds ago and made no changes, the cache lets you skip.
 
+## Enforcement Gates
+
+Prompt-level rules get ignored under pressure. These gates are **hard blocks** — they must be checked at the tool/action level, not just remembered.
+
+### Gate 1: Completion Claims Require verify.log
+
+Before ANY completion claim (bead close, PR creation, `/ship`, task completion):
+
+1. Check `.beads/verify.log` exists and contains a recent `PASS` stamp
+2. If verify.log is missing or stale (older than last file change) → **BLOCK** — run verification first
+3. If verify.log shows `FAIL` → **BLOCK** — do not proceed
+
+```
+✅ verify.log exists, PASS within last edit window → proceed
+❌ verify.log missing → STOP: "Run verification first"
+❌ verify.log shows FAIL → STOP: "Verification failed, fix before claiming complete"
+❌ verify.log stale (files changed since last PASS) → STOP: "Re-run verification"
+```
+
+### Gate 2: Agent Delegation Requires Post-Verification
+
+After ANY `task()` subagent returns with "success", follow the **Worker Distrust Protocol** from AGENTS.md — read changed files, run verification, check acceptance criteria. Do not trust agent self-reports.
+
+### Enforcement Principle
+
+> **Prompt rules fail under pressure. Gates fail safe.**
+>
+> When a constraint matters enough to be an iron law, enforce it at the action level:
+> check a file, verify a condition, reject if unmet. Don't rely on the agent
+> "remembering" to follow the rule.
+
 ## Why This Matters
 
 From 24 failure memories:

package/package.json

@@ -1,6 +1,6 @@
 {
 	"name": "opencodekit",
-	"version": "0.18.6",
+	"version": "0.18.8",
 	"description": "CLI tool for bootstrapping and managing OpenCodeKit projects",
 	"keywords": [
 		"agents",