claude-dev-env 2.8.0 → 2.10.0

package/skills/team-advisor/SKILL.md

@@ -2,63 +2,42 @@
 name: team-advisor
 description: >-
   Spawn one warm session-advisor at the strongest reachable tier and consult
-  it before big decisions, completion, commits, or when stuck. Triggers:
+  it before substantive work, completion, commits, or when stuck. Triggers:
   'team-advisor', 'team advisor', 'second opinion', 'advisor', 'consult',
   'verify', 'validate', 'commit', 'push'.
 ---
 
 # Team Advisor
 
-## Principle
+One warm advisor at the strongest tier this session can reach. This session is the sole consumer.
 
-One warm, addressable advisor available at the strongest model tier the session can reach. The session sends concise briefs when a decision benefits from a second opinion: before acting on a plan, at completion, before commits, when stuck, when reconsidering the approach, or when an agent deems it necessary and beneficial to the user's goals.
+## Refs
 
-## Follow the shared protocol
-
-**Detect the host profile first** (Host profiles in
-[`_shared/advisor/advisor-protocol.md`](../../_shared/advisor/advisor-protocol.md)
-— e.g. %USERPROFILE%\.claude\_shared\advisor\advisor-protocol.md). Do not start a model-floor
-walk until the host is known. 
-
-This session is the shared advisor's sole consumer, so its model floor is
-simply this session's own tier — no routing table to take a max against.
+| Doc | Holds |
+|---|---|
+| [`docs/references/advisor-tool.md`](../../docs/references/advisor-tool.md) | **Consult cadence and weight** — when to call, hard rule before first write, how to treat advice. Read this for every consult. |
+| [`~/.claude/_shared/advisor/advisor-protocol.md`](../../_shared/advisor/advisor-protocol.md) | **Bind and lifecycle** — host detect, model floor, warm-up or CLI bind, charter, drift re-bind, CLI fallback, executor paste blocks. |
+| [`agents/session-advisor.md`](../../agents/session-advisor.md) | **Reply contract** — ENDORSE / CORRECTION / PLAN / STOP; SendMessage only. |
 
-**Claude host:** follow the shared protocol for the model-floor walk, the
-warm-up spawn and charter, the consult format and cadence, drift-respawn, and
-the CLI fallback — using `team-advisor-agent` as the name and this session as
-the only consumer (skip the "who you are and your assignment" opener in each
-consult; a single-consumer session doesn't need it). A **Fable**-tier attempt
-carries the exact token `FABLE-SPAWN-AUTHORIZED` in its spawn prompt, as the
-protocol's warm-up rule states; `hooks/blocking/fable_spawn_gate.py` denies a
-fable spawn whose prompt lacks it.
+## Bind
 
-**Third-party host:** bind a max-tier Claude advisor through the shared CLI Claude-chain
-in the protocol (Fable max, then Opus max; `claude_chain_runner.py` ranks
-`~/.claude/claude-chain.json` accounts by weekly remaining via `claude_chain_usage`
-and fails over on usage limits). Consult via
-`--resume <session_id>` on that bind. This session is the sole consumer of that
-CLI advisor; skip the multi-consumer opener. When the chain cannot bind or
-reply, fail closed and report to the user — do **not** answer ENDORSE /
-CORRECTION / PLAN / STOP as this third-party session.
+1. Detect the host profile first (protocol **Host profiles**), then walk the model floor.
+2. Floor: this session's own tier on Claude; Opus floor with Fable first on a third-party host.
+3. Name: `team-advisor-agent` on Claude (Agent spawn of `session-advisor`); one CLI `session_id` on a third-party host via the protocol Claude-chain.
+4. A Fable-tier spawn or re-spawn carries the exact token `FABLE-SPAWN-AUTHORIZED` in its prompt (protocol warm-up; `fable_spawn_gate` requires it).
+5. Skip the multi-consumer "who you are" opener — sole consumer.
+6. When the bind or reply path fails, fail closed and report to the user. On a third-party host, only the bound Claude advisor issues ENDORSE / CORRECTION / PLAN / STOP.
 
-## Constraints
+Full walk, charter, consult message shape, and drift re-bind live in the protocol.
 
-- One advisor bind per session (`team-advisor-agent` on Claude; one CLI
-  `session_id` on a third-party host), owned by this session for its whole lifecycle
-  (spawn or CLI bind, drift re-bind, shutdown) — see
-  [`_shared/advisor/advisor-protocol.md`](../../_shared/advisor/advisor-protocol.md).
-- Never bind the advisor, or its CLI path, at a tier below the protocol floor
-  for this host (Claude: this session's own tier; a third-party host: Opus floor
-  with Fable first).
-- The advisor only answers. It never edits a file, never runs a build or
-  test, and never posts anything on the session's behalf.
+## Consult
 
-## File Index
+Follow **When to call**, **Hard rule**, and **How to treat advice** in `advisor-tool.md`.
 
-| File | Purpose |
-|---|---|
-| `SKILL.md` | Pointer to the shared advisor protocol; this session's consumer-specific wiring and constraints. |
+Each brief: delta since last consult, live decision or blocker, paths or excerpts needed. Protocol owns the full consult format.
 
-## Folder Map
+## Constraints
 
-- `SKILL.md` — complete team-advisor workflow instructions.
+- One bind per session; this session owns spawn or CLI bind, drift re-bind, and shutdown.
+- Bind at or above the protocol floor for this host.
+- The advisor only answers (messaging); the session runs tools and posts.

package/system-prompts/software-engineer.xml

@@ -28,8 +28,8 @@
   Lead every response with the direct answer or result. Open with the substance — the
   conclusion, the fix, the decision — before any supporting context or explanation. Expand
   depth in proportion to the complexity of the ask: a direct question earns a sentence; a
-  multi-file investigation earns structured explanation. Depth scales with the ask; a fixed
-  reply-length cap is wrong.
+  multi-file investigation or requested full audit earns structured explanation long enough
+  to cover the substance. Depth scales with the ask; a fixed reply-length cap is wrong.
 
   Reference specific code locations using the pattern `file_path:line_number` so the user can
   navigate directly to the source.
@@ -90,8 +90,10 @@
 </problem_solving>
 
 <code_quality>
-  Apply the system-prompt code-quality checklist as the canonical naming and style source for
-  all code you write or modify. ~/.claude/docs/CODE_RULES.md is a pointer to this source.
+  Canonical code-quality policy is repository-root AGENTS.md (installed review contract).
+  ~/.claude/docs/CODE_RULES.md is the compact projection of AGENTS.md for generation load —
+  when the two disagree, AGENTS.md wins. The checklist below is a short reminder, not a
+  second canonical source.
 
   Naming — full words only:
   - Use "context" over "ctx", "configuration" over "cfg", "message" over "msg",
@@ -109,7 +111,8 @@
   Types — complete coverage required:
   - Every parameter and return value carries an explicit type hint
   - Replace Any with the most specific type available
-  - Replace every # type: ignore annotation with correct typing
+  - Prefer a real type over `# type: ignore`; when an ignore is required, add a second
+    trailing `#` justification of at least five characters (AGENTS Types)
 
   Configuration — one source of truth:
   - Every constant lives in config/ only
@@ -378,6 +381,15 @@
   work around tool limitations — are removed when the task is done.
 </cleanup>
 
+<communication_contract>
+  Visible output is concise by default. The first progress update is one sentence.
+  Later updates mark only important discoveries or direction changes. The final
+  response begins with the outcome. When thinking is disabled, speak one brief
+  natural-language sentence before a tool call that needs context, state when no
+  fitting tool exists, and never emit internal-system XML in visible output.
+  Full contract: rules/opus5-communication-contract.md (opus5-communication-contract-v1).
+</communication_contract>
+
 <help_and_feedback>
   When the user asks for help with this tool or wants to report a problem:
   - For usage help, direct them to: /help

package/commands/initialize.md

@@ -1,90 +0,0 @@
-# Initialize Session
-
-You are starting a new task. Before proceeding, you MUST read and internalize the critical protocols from CLAUDE.md.
-
-## MANDATORY PROTOCOL REVIEW
-
-Read CLAUDE.md now and focus on:
-
-### 1. AGENT-FIRST WORKFLOW (CRITICAL)
-
-**BEFORE RESPONDING TO ANY USER MESSAGE:**
-- ☐ What is the user asking for?
-- ☐ Does ANY agent match this request?
-- ☐ If yes → INVOKE THE AGENT via Task tool
-- ☐ If no agent matches → Proceed with skills/direct implementation
-
-**Agent Decision Tree:**
-- Automation work → check available agents for automation patterns
-- Web App Development → check available agents for framework-specific patterns
-
-**Never skip agent check to "save time" or because request "seems simple"**
-
-### 2. SKILLS WORKFLOW (CRITICAL)
-
-**Before ANY task:**
-- ☐ List available skills in your mind
-- ☐ Does ANY skill match this request?
-- ☐ If yes → Use Skill tool to read and run it
-- ☐ Announce which skill you're using
-- ☐ Follow the skill exactly
-
-**Mandatory skills to remember:**
-- `code-standards` - For ALL code generation, planning, implementation
-- `superpowers:test-driven-development` - For ANY feature/bugfix
-- `superpowers:brainstorming` - BEFORE coding on design tasks
-- `superpowers:systematic-debugging` - For bugs/failures
-- `superpowers:verification-before-completion` - Before claiming work is done
-
-### 3. TDD IS NON-NEGOTIABLE
-
-**Red-Green-Refactor:**
-1. **Red**: Write failing test FIRST. NO production code.
-2. **Green**: MINIMUM code to pass test only.
-3. **Refactor**: Assess improvements after green.
-
-**If writing production code without failing test, STOP immediately.**
-
-### 4. CODE STANDARDS (ENFORCED)
-
-- No `Any` types - Use Union/Optional/generics
-- No `# type: ignore` without justification
-- All imports at top
-- No files in project root
-- Immutable data (frozen dataclasses)
-- Small functions (5-15 lines, max 30)
-- Max 2 nesting levels
-- No comments (self-documenting code)
-- No Unicode in print/debug (Windows compatibility)
-
-### 5. RIGHT-SIZED ENGINEERING
-
-**Always Do:**
-- Extract constants and configuration
-- Create reusable functions
-- DRY from the start
-- Single responsibility
-
-**Never Do (Solo Scale):**
-- Abstract base classes for single implementations
-- Dependency injection frameworks
-- Complex patterns
-- Over-abstracted interfaces
-
-## Common Rationalizations That Mean Failure
-
-If you think:
-- "This is too simple" → WRONG. Check agents/skills.
-- "Let me just implement quickly" → WRONG. Check agents/skills first.
-- "I'll gather context first" → WRONG. Agents/skills tell you HOW.
-- "This doesn't need formal process" → WRONG. Process prevents mistakes.
-
-## NOW PROCEED
-
-You have reviewed the critical protocols. Apply them to the current task.
-
-**Remember:**
-1. Check agents FIRST
-2. Check skills SECOND
-3. Follow TDD ALWAYS
-4. Build it right, but build it simple

package/commands/stubcheck.md

@@ -1,88 +0,0 @@
----
-description: Detect and fix stub implementations in code and plans with agent/skill recommendations
-allowed-tools: Read, Grep, Glob, Skill
----
-
-Execute comprehensive stub detection using the stub-detector skill to identify incomplete implementations and recommend appropriate solutions.
-
-## Process
-
-1. **Invoke stub-detector skill** to scan current project
-2. Identify all stubs in:
-   - Python code (pass, NotImplementedError, TODO comments)
-   - Plan documents (incomplete steps, TBD markers)
-   - Configuration (placeholder values)
-3. Classify by severity (CRITICAL/HIGH/MEDIUM/LOW)
-4. Recommend specific agents or skills based on:
-   - Project type (Web Automation, Django, etc.)
-   - Stub category (config, automation, business logic)
-   - Available implementations (check for existing solutions first)
-5. Present actionable report with:
-   - file:line references for each stub
-   - Severity and impact assessment
-   - Specific recommendations with rationale
-   - Batch operation opportunities
-6. Offer immediate actions to fix CRITICAL stubs
-
-## Stub Detection Patterns
-
-**Code stubs to find:**
-```
-pass statements
-NotImplementedError
-TODO/FIXME/HACK comments
-Empty functions with only docstrings
-Placeholder returns (return None when type suggests otherwise)
-```
-
-**Documentation stubs to find:**
-```
-[TODO] or [TBD] markers
-"To be implemented" sections
-Incomplete step descriptions
-Placeholder text in plans
-Missing details in implementation guides
-```
-
-## Agent/Skill Recommendation Logic
-
-**Web Automation stubs →**
-- Automation: automation-agent or automation skill
-- Google Sheets: sheets orchestrator agent
-- Config issues: clean-coder agent (extracts magic values to `config/` per CODE_RULES.md)
-
-**Web framework stubs →**
-- Models/views/forms: domain-specific agent
-- Business logic: domain-specific agent
-- Domain features: domain-specific agent
-
-**General stubs →**
-- Follow TDD workflow from CLAUDE.md
-- Follow CODE_RULES.md standards (via clean-coder agent)
-- Check for existing shared utilities first
-
-## Output Format
-
-Provide a clear summary report showing:
-- Total stubs found with severity breakdown
-- Grouped by category for batch processing
-- Specific recommendations for each stub
-- Immediate action options for user
-
-Example:
-```
-Found 7 stubs (3 CRITICAL, 2 HIGH, 2 MEDIUM)
-
-[CRITICAL] services/file_processor.py:45
-Incomplete matching logic
-→ Recommendation: Use FileProcessor (already available)
-→ Action: Refactor to use shared utility
-
-[HIGH] tests/test_integration.py:234
-Missing integration test
-→ Recommendation: Follow TDD workflow
-
-Would you like me to fix CRITICAL stubs now?
-```
-
-**IMPORTANT:** Always check if solutions already exist (like we just implemented FileProcessor) before recommending new development.

package/skills/test_markdown_link_integrity.py

@@ -1,107 +0,0 @@
-"""Link-integrity check for the skills tree and shared protocol docs.
-
-Walks every markdown file under ``skills/``, ``_shared/pr-loop/``, and
-``_shared/advisor/`` and resolves each relative markdown link target against
-the linking file's directory. A link whose target file or directory does not
-exist on disk is reported with its source file and line number, and the test
-fails.
-"""
-
-from __future__ import annotations
-
-import re
-from pathlib import Path
-
-
-SKILLS_ROOT = Path(__file__).parent
-SHARED_PR_LOOP_ROOT = SKILLS_ROOT.parent / "_shared" / "pr-loop"
-SHARED_ADVISOR_ROOT = SKILLS_ROOT.parent / "_shared" / "advisor"
-
-MARKDOWN_LINK_PATTERN = re.compile(r"\]\(([^)\s]+)\)")
-FENCE_MARKER = "```"
-
-SKIPPED_TARGET_PREFIXES = (
-    "http://",
-    "https://",
-    "mailto:",
-    "#",
-    "/",
-    "$",
-    "~",
-    "<",
-)
-
-
-def _iter_markdown_files() -> list[Path]:
-    """List every markdown file the integrity check covers.
-
-    Returns:
-        All ``.md`` files under the skills tree, the shared pr-loop tree, and
-        the shared advisor tree.
-    """
-    all_markdown_files = sorted(SKILLS_ROOT.rglob("*.md"))
-    all_markdown_files.extend(sorted(SHARED_PR_LOOP_ROOT.rglob("*.md")))
-    all_markdown_files.extend(sorted(SHARED_ADVISOR_ROOT.rglob("*.md")))
-    return all_markdown_files
-
-
-def _is_checkable_target(link_target: str) -> bool:
-    """Decide whether a markdown link target names a local relative path.
-
-    Args:
-        link_target: The raw text between ``](`` and ``)`` in a markdown link.
-
-    Returns:
-        True when the target is a relative filesystem path this check can
-        resolve; False for URLs, anchors, absolute paths, environment-variable
-        paths, home-relative paths, and angle-bracket placeholders.
-    """
-    if link_target.startswith(SKIPPED_TARGET_PREFIXES):
-        return False
-    if "\\" in link_target and link_target.count(":") > 0:
-        return False
-    return True
-
-
-def _collect_broken_links(markdown_path: Path) -> list[str]:
-    """Resolve every relative link in one markdown file.
-
-    Fenced code blocks are skipped, and a ``#fragment`` suffix is stripped
-    before resolution.
-
-    Args:
-        markdown_path: The markdown file whose links are resolved.
-
-    Returns:
-        One ``file:line -> target`` description per unresolvable link.
-    """
-    all_broken_links: list[str] = []
-    is_inside_fence = False
-    for line_number, each_line in enumerate(
-        markdown_path.read_text(encoding="utf-8").splitlines(), start=1
-    ):
-        if each_line.lstrip().startswith(FENCE_MARKER):
-            is_inside_fence = not is_inside_fence
-            continue
-        if is_inside_fence:
-            continue
-        for each_match in MARKDOWN_LINK_PATTERN.finditer(each_line):
-            link_target = each_match.group(1).split("#", 1)[0]
-            if not link_target or not _is_checkable_target(link_target):
-                continue
-            resolved_target = (markdown_path.parent / link_target).resolve()
-            if not resolved_target.exists():
-                all_broken_links.append(
-                    f"{markdown_path}:{line_number} -> {each_match.group(1)}"
-                )
-    return all_broken_links
-
-
-def test_every_relative_markdown_link_resolves() -> None:
-    all_broken_links: list[str] = []
-    for each_markdown_path in _iter_markdown_files():
-        all_broken_links.extend(_collect_broken_links(each_markdown_path))
-    broken_link_report = "\n".join(all_broken_links)
-    assert not all_broken_links, (
-        f"Unresolvable relative markdown links:\n{broken_link_report}"
-    )