devlyn-cli 2.5.3 → 2.5.4

package/benchmark/auto-resolve/scripts/judge-opus-pass.sh

@@ -1,7 +1,7 @@
 #!/usr/bin/env bash
-# judge-opus-pass.sh — POST-RUN Opus 4.7 dual-judge sidecar.
+# judge-opus-pass.sh — POST-RUN Opus dual-judge sidecar.
 #
-# Runs Opus 4.7 over the SAME sanitized blind prompts that judge.sh wrote
+# Runs Opus (the Claude CLI's default model) over the SAME sanitized blind prompts that judge.sh wrote
 # (judge-prompt.txt), producing a parallel judge-opus.json alongside the
 # canonical judge.json. Does NOT modify judge.sh and does NOT change the
 # measurement contract for the in-flight run — this is a sidecar that adds
@@ -107,7 +107,7 @@ PY
 command -v claude >/dev/null 2>&1 || { echo "claude CLI not on PATH"; exit 1; }
 
 CLAUDE_CLI_VER=$(claude --version 2>/dev/null || echo "claude-cli unknown")
-JUDGE_MODEL_ALIAS="opus-4.7"
+JUDGE_MODEL_ALIAS="opus-default"
 
 echo "[opus-judge] run-id=$RUN_ID  cli=$CLAUDE_CLI_VER  model=$JUDGE_MODEL_ALIAS"
 

package/config/skills/_shared/adapters/README.md

@@ -14,10 +14,10 @@ This is also the load-bearing piece for **multi-LLM evolution**. When Qwen / Gem
 
 ## Format
 
-Each adapter is a single markdown file named `<model-id>.md` (e.g. `opus-4-7.md`, `gpt-5-5.md`). Structure:
+Each adapter is a single markdown file named `<engine>.md`, matching the `--engine` selector value (e.g. `claude.md`, `codex.md`). Structure:
 
 ```markdown
-# <Model name> adapter
+# <Engine> adapter
 
 > Source: <official-prompt-engineering-guide URL>
 

package/config/skills/_shared/adapters/{opus-4-7.md → claude.md}

@@ -1,10 +1,10 @@
-# Claude Opus 4.7 adapter
+# Claude Opus adapter
 
 > Source: <https://platform.claude.com/docs/en/build-with-claude/prompt-engineering/claude-prompting-best-practices>
 
 ## Identity
 
-You are Claude Opus 4.7 by Anthropic. Anthropic's prompt-engineering guide for this model governs your behavior on top of the canonical phase prompt below. When the canonical body and this header conflict on tactics, the canonical body wins on what to deliver; this header wins on how to deliver it.
+You are Claude Opus by Anthropic. Anthropic's prompt-engineering guide for this model governs your behavior on top of the canonical phase prompt below. When the canonical body and this header conflict on tactics, the canonical body wins on what to deliver; this header wins on how to deliver it.
 
 ## Output discipline
 
@@ -16,7 +16,7 @@ When prompt maintenance adds examples for Claude, prefer concise positive exampl
 
 ## Tool-use posture
 
-You default to fewer tool calls than prior Claude generations. When the canonical body lists tools, use them when their result would change your answer. Make independent tool calls in parallel; chain only when one depends on another's output. Do not narrate "I'll now call X" preambles unless the canonical body requests progress updates.
+When the canonical body lists tools, use them when their result would change your answer. Make independent tool calls in parallel; chain only when one depends on another's output. Do not narrate "I'll now call X" preambles unless the canonical body requests progress updates.
 
 ## Effort and autonomy
 
@@ -28,10 +28,10 @@ When the canonical body asks you to verify your output before declaring done ("s
 
 ## Anti-patterns
 
-You interpret instructions more literally than prior Claude versions. The official guide is explicit about three failure modes:
+You interpret instructions literally and explicitly. The official guide is explicit about three failure modes:
 
 1. **Review-prompt self-filtering**: when the canonical body asks for findings, report every issue you find — including low-severity and low-confidence ones; do not filter for importance or confidence. The harness has a separate filter step.
-2. **Subagent over-spawning**: do NOT spawn a subagent for work you can complete in a single response. Spawn only when the canonical body explicitly requests it OR when fanning out across independent items.
+2. **Subagent spawning**: do NOT spawn a subagent for work you can complete in a single response. Spawn only when the canonical body explicitly requests it OR when fanning out across independent items.
 3. **Overengineering**: do NOT add files, abstractions, error handling, validation, or "future flexibility" beyond what the spec asks. A bug fix doesn't need surrounding cleanup. The right complexity is the minimum needed for the current task.
 
 You do NOT need stronger imperatives ("CRITICAL!", "YOU MUST!") to follow rules. Normal phrasing is sufficient.

package/config/skills/_shared/adapters/{gpt-5-5.md → codex.md}

@@ -1,10 +1,10 @@
-# OpenAI GPT-5.5 adapter
+# Codex (OpenAI) adapter
 
 > Source: <https://developers.openai.com/api/docs/guides/prompt-guidance?model=gpt-5.5>
 
 ## Identity
 
-You are GPT-5.5 by OpenAI. OpenAI's prompt-guidance for this model governs your behavior on top of the canonical phase prompt below. When the canonical body and this header conflict on tactics, the canonical body wins on what to deliver; this header wins on how to deliver it.
+You are OpenAI's Codex model. OpenAI's prompt-guidance for this model governs your behavior on top of the canonical phase prompt below. When the canonical body and this header conflict on tactics, the canonical body wins on what to deliver; this header wins on how to deliver it.
 
 ## Output discipline
 
@@ -28,6 +28,6 @@ The official guide warns explicitly about carrying over instructions from older
 
 ## Prompt-maintenance cue
 
-When asked to improve a failed prompt, act as GPT-5.5 metaprompter for itself: name the observed failure, then propose the smallest instruction to add, remove, or relocate. Prefer subtractive changes before adding new rules; keep the canonical body model-neutral and put only GPT-specific tactics in this adapter.
+When asked to improve a failed prompt, act as a metaprompter for itself: name the observed failure, then propose the smallest instruction to add, remove, or relocate. Prefer subtractive changes before adding new rules; keep the canonical body model-neutral and put only GPT-specific tactics in this adapter.
 
 Do not narrate internal deliberation. State results and decisions directly.

package/config/skills/devlyn:ideate/SKILL.md

@@ -28,7 +28,7 @@ Read `_shared/runtime-principles.md` (Subtractive-first / Goal-locked / No-worka
 </harness_principles>
 
 <engine_routing>
-Default engine: Claude. The per-engine adapter from `_shared/adapters/<model>.md` is prepended to the elicitation prompt so the model honors its own official prompt-engineering guidance during the Q&A.
+Default engine: Claude. The per-engine adapter from `_shared/adapters/<engine>.md` is prepended to the elicitation prompt so the model honors its own official prompt-engineering guidance during the Q&A.
 </engine_routing>
 
 <modes>

package/config/skills/devlyn:resolve/SKILL.md

@@ -28,7 +28,7 @@ Every phase reads `_shared/runtime-principles.md` (Subtractive-first / Goal-lock
 </harness_principles>
 
 <engine_routing>
-Each phase routes to an engine and prepends the per-engine adapter header from `_shared/adapters/<model>.md` to the canonical phase body. Adapter is the per-model delta (Anthropic Opus 4.7 guide for Claude, OpenAI GPT-5.5 guide for Codex). Canonical body is engine-agnostic.
+Each phase routes to an engine and prepends the per-engine adapter header from `_shared/adapters/<engine>.md` (e.g. `claude.md`, `codex.md`) to the canonical phase body. Adapter is the per-model delta (Anthropic's Opus prompt-engineering guide for Claude, OpenAI's GPT-5.5 guide for Codex). Canonical body is engine-agnostic.
 
 - Claude phases: spawn `Agent` (`mode: "bypassPermissions"`); prompt = adapter-header + canonical-body + task-context.
 - Codex phases: shell out via `bash _shared/codex-monitored.sh` with the same compounded prompt. The wrapper closes stdin and emits a heartbeat. No MCP.

package/config/skills/devlyn:resolve/references/phases/plan.md

@@ -1,6 +1,6 @@
 # PHASE 1 — PLAN (canonical body)
 
-The per-engine adapter header from `_shared/adapters/<model>.md` is prepended at runtime. This file is engine-agnostic.
+The per-engine adapter header from `_shared/adapters/<engine>.md` is prepended at runtime. This file is engine-agnostic.
 
 <role>
 You translate a spec or generated criteria into a concrete plan: the file list to touch, the risks the implementation must navigate, and a verbatim restatement of what acceptance requires. The plan is the contract IMPLEMENT executes against.

package/optional-skills/prompt-engineering/SKILL.md

@@ -5,7 +5,7 @@ allowed-tools: Read, Grep, Glob, Edit, Write
 ---
 
 <!--
-  Version: 2026-01 | Claude 4.x (Opus 4.5, Sonnet 4.5, Haiku 4.5)
+  Version: 2026-05 | Claude 4.x (Opus 4.8, Sonnet 4.6, Haiku 4.5)
   Source: https://platform.claude.com/docs/build-with-claude/prompt-engineering/claude-4-best-practices
 -->
 
@@ -103,13 +103,13 @@ explicitly requested. Incorporate information naturally into sentences.
 START: Which Claude model?
     │
     ├─► Complex reasoning, multi-step analysis, or highest quality?
-    │       → Opus 4.5 (claude-opus-4-5-20250929)
+    │       → Opus 4.8 (claude-opus-4-8)
     │
     ├─► Balance of quality, speed, and cost?
-    │       → Sonnet 4.5 (claude-sonnet-4-5-20250929)
+    │       → Sonnet 4.6 (claude-sonnet-4-6)
     │
     └─► Fast responses, simple tasks, cost-sensitive?
-            → Haiku 4.5 (claude-haiku-4-5-20250929)
+            → Haiku 4.5 (claude-haiku-4-5-20251001)
 ```
 
 ---
@@ -152,7 +152,7 @@ results, execute sequentially. Never use placeholders for missing parameters.
 |--------------|--------------|-----|
 | "Can you suggest changes?" | Claude will only suggest, not implement | "Change this function to..." |
 | "Think about this carefully" | Triggers thinking sensitivity | "Consider this carefully" |
-| "CRITICAL: You MUST..." | Opus 4.5 may over-trigger | "Use this when..." |
+| "CRITICAL: You MUST..." | Claude may over-trigger | "Use this when..." |
 | Vague instructions | Literal interpretation | Add explicit detail |
 | No context for rules | Claude can't generalize | Explain WHY behind rules |
 | Examples with unwanted patterns | Claude mimics examples precisely | Remove unwanted behaviors from examples |

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "devlyn-cli",
-  "version": "2.5.3",
+  "version": "2.5.4",
   "description": "AI development toolkit for Claude Code — ideate, resolve, and ship with context engineering and agent orchestration",
   "homepage": "https://github.com/fysoul17/devlyn-cli#readme",
   "bin": {

package/scripts/lint-skills.sh

@@ -75,8 +75,8 @@ devlyn:resolve/references/phases/cleanup.md
 devlyn:resolve/references/phases/verify.md
 _shared/expected.schema.json
 _shared/adapters/README.md
-_shared/adapters/opus-4-7.md
-_shared/adapters/gpt-5-5.md
+_shared/adapters/claude.md
+_shared/adapters/codex.md
 _shared/codex-config.md
 _shared/codex-monitored.sh
 _shared/engine-preflight.md
@@ -1586,29 +1586,29 @@ fi
 section "Check 10d: Prompt adapters cite official model guidance"
 adapter_missing=0
 if ! grep -Fq 'https://developers.openai.com/api/docs/guides/prompt-guidance?model=gpt-5.5' \
-  config/skills/_shared/adapters/gpt-5-5.md; then
-  bad "gpt-5-5 adapter missing official OpenAI prompt guidance URL"
+  config/skills/_shared/adapters/codex.md; then
+  bad "codex adapter missing official OpenAI prompt guidance URL"
   adapter_missing=1
 fi
 if ! grep -Fq 'https://platform.claude.com/docs/en/build-with-claude/prompt-engineering/claude-prompting-best-practices' \
-  config/skills/_shared/adapters/opus-4-7.md; then
-  bad "opus-4-7 adapter missing official Claude prompting best-practices URL"
+  config/skills/_shared/adapters/claude.md; then
+  bad "claude adapter missing official Claude prompting best-practices URL"
   adapter_missing=1
 fi
 for pattern in 'Use Markdown only where it carries structure' 'metaprompter for itself'; do
-  if ! grep -Fq "$pattern" config/skills/_shared/adapters/gpt-5-5.md; then
-    bad "gpt-5-5 adapter missing official-guidance tactic: $pattern"
+  if ! grep -Fq "$pattern" config/skills/_shared/adapters/codex.md; then
+    bad "codex adapter missing official-guidance tactic: $pattern"
     adapter_missing=1
   fi
 done
 for pattern in 'high` or `xhigh` effort' 'report every issue you find' 'do not filter for importance or confidence' 'prefer concise positive examples' '<example>'; do
-  if ! grep -Fq "$pattern" config/skills/_shared/adapters/opus-4-7.md; then
-    bad "opus-4-7 adapter missing official-guidance tactic: $pattern"
+  if ! grep -Fq "$pattern" config/skills/_shared/adapters/claude.md; then
+    bad "claude adapter missing official-guidance tactic: $pattern"
     adapter_missing=1
   fi
 done
 for file in config/skills/devlyn:resolve/SKILL.md config/skills/devlyn:ideate/SKILL.md; do
-  if ! grep -Fq '_shared/adapters/<model>.md' "$file"; then
+  if ! grep -Fq '_shared/adapters/<engine>.md' "$file"; then
     bad "$file — missing per-engine adapter injection contract"
     adapter_missing=1
   fi