dino-spec 18.3.0 → 18.5.0

package/.claude/rules/dino/auto-context-engineering.md

@@ -1,44 +1,44 @@
-# Auto Context Engineering (dino v18.2.0)
-
-## Automatic Features
-
-All context engineering features are enabled by default:
-
-1. **MCP Tool Filtering** - Only phase-relevant tools loaded (~50-70% context savings)
-2. **Lazy Context Loading** - Resources deferred until needed
-3. **Session State Tracking** - Focus, blockers, progress auto-tracked
-
-## No Configuration Required
-
-Users do NOT need to:
-- Set environment variables
-- Pass special flags
-- Configure anything manually
-
-These features auto-apply via `dino init` and `dino update`.
-
-## Opt-Out (Not Recommended)
-
-To disable MCP tool filtering:
-```bash
-DINO_AGGRESSIVE_FILTERING=false
-```
-
-This is NOT recommended as it increases context overhead significantly.
-
-## Token Savings by Phase
-
-| Phase | Tools Loaded | Savings |
-|-------|-------------|---------|
-| Discovery | ~17 | ~60% |
-| Planning | ~22 | ~53% |
-| Implementation | ~23 | ~51% |
-| Completing | ~12 | ~74% |
-
-## Phase Auto-Detection
-
-Phase is auto-detected from:
-1. `## Phase` section in session.md
-2. Default: `discovery` if not set
-
-The MCP server parses session.md on each `tools/list` request to get the current phase.
+# Auto Context Engineering (dino v18.2.0)
+
+## Automatic Features
+
+All context engineering features are enabled by default:
+
+1. **MCP Tool Filtering** - Only phase-relevant tools loaded (~50-70% context savings)
+2. **Lazy Context Loading** - Resources deferred until needed
+3. **Session State Tracking** - Focus, blockers, progress auto-tracked
+
+## No Configuration Required
+
+Users do NOT need to:
+- Set environment variables
+- Pass special flags
+- Configure anything manually
+
+These features auto-apply via `dino init` and `dino update`.
+
+## Opt-Out (Not Recommended)
+
+To disable MCP tool filtering:
+```bash
+DINO_AGGRESSIVE_FILTERING=false
+```
+
+This is NOT recommended as it increases context overhead significantly.
+
+## Token Savings by Phase
+
+| Phase | Tools Loaded | Savings |
+|-------|-------------|---------|
+| Discovery | ~17 | ~60% |
+| Planning | ~22 | ~53% |
+| Implementation | ~23 | ~51% |
+| Completing | ~12 | ~74% |
+
+## Phase Auto-Detection
+
+Phase is auto-detected from:
+1. `## Phase` section in session.md
+2. Default: `discovery` if not set
+
+The MCP server parses session.md on each `tools/list` request to get the current phase.

package/.claude/rules/dino/change-categories.md

@@ -1,60 +1,60 @@
-# Change Categories (dino v18.2.0)
-
-## File Classification
-
-Files are classified into categories based on path and name:
-
-### dependency (High Significance)
-Files that affect project dependencies:
-- `package.json`, `package-lock.json`, `yarn.lock`, `pnpm-lock.yaml`
-- `requirements.txt`, `Pipfile`, `pyproject.toml`
-- `Cargo.toml`, `Cargo.lock`
-- `go.mod`, `go.sum`
-- `Gemfile`, `Gemfile.lock`
-
-**Pending Action:** Run package manager install command
-
-### structural (High Significance)
-Files that affect project structure:
-- Entry points: `index.ts`, `main.ts`, `app.ts`
-- Configuration: `tsconfig.json`, `webpack.config.js`
-- Build configs: `build.js`, `vite.config.ts`
-- Directory additions/removals
-
-**Pending Action:** Verify import paths
-
-### config (Medium Significance)
-Configuration files:
-- `.env`, `.env.*`
-- `*.config.js`, `*.config.ts`
-- `.eslintrc`, `.prettierrc`
-- CI/CD files: `.github/`, `.gitlab-ci.yml`
-
-### test (Medium Significance)
-Test files:
-- `*.test.ts`, `*.test.js`
-- `*.spec.ts`, `*.spec.js`
-- `test/`, `tests/`, `__tests__/` directories
-
-### code (Low Significance)
-Source code files:
-- `src/`, `lib/` directories
-- Main programming language files
-- Utility and helper files
-
-### documentation (Low Significance)
-Documentation files:
-- `*.md`, `README*`
-- `docs/` directory
-- Code comments (when detected)
-
-## Impact on Workflow
-
-| Category | Run Tests | Run Build | Review |
-|----------|-----------|-----------|--------|
-| dependency | After install | Yes | Changelog |
-| structural | Yes | Yes | Imports |
-| config | Maybe | Maybe | Side effects |
-| test | Yes | No | Coverage |
-| code | Yes | Yes | Normal |
-| documentation | No | No | Spelling |
+# Change Categories (dino v18.2.0)
+
+## File Classification
+
+Files are classified into categories based on path and name:
+
+### dependency (High Significance)
+Files that affect project dependencies:
+- `package.json`, `package-lock.json`, `yarn.lock`, `pnpm-lock.yaml`
+- `requirements.txt`, `Pipfile`, `pyproject.toml`
+- `Cargo.toml`, `Cargo.lock`
+- `go.mod`, `go.sum`
+- `Gemfile`, `Gemfile.lock`
+
+**Pending Action:** Run package manager install command
+
+### structural (High Significance)
+Files that affect project structure:
+- Entry points: `index.ts`, `main.ts`, `app.ts`
+- Configuration: `tsconfig.json`, `webpack.config.js`
+- Build configs: `build.js`, `vite.config.ts`
+- Directory additions/removals
+
+**Pending Action:** Verify import paths
+
+### config (Medium Significance)
+Configuration files:
+- `.env`, `.env.*`
+- `*.config.js`, `*.config.ts`
+- `.eslintrc`, `.prettierrc`
+- CI/CD files: `.github/`, `.gitlab-ci.yml`
+
+### test (Medium Significance)
+Test files:
+- `*.test.ts`, `*.test.js`
+- `*.spec.ts`, `*.spec.js`
+- `test/`, `tests/`, `__tests__/` directories
+
+### code (Low Significance)
+Source code files:
+- `src/`, `lib/` directories
+- Main programming language files
+- Utility and helper files
+
+### documentation (Low Significance)
+Documentation files:
+- `*.md`, `README*`
+- `docs/` directory
+- Code comments (when detected)
+
+## Impact on Workflow
+
+| Category | Run Tests | Run Build | Review |
+|----------|-----------|-----------|--------|
+| dependency | After install | Yes | Changelog |
+| structural | Yes | Yes | Imports |
+| config | Maybe | Maybe | Side effects |
+| test | Yes | No | Coverage |
+| code | Yes | Yes | Normal |
+| documentation | No | No | Spelling |

package/.claude/rules/dino/effort-guidance.md

@@ -1,109 +1,109 @@
-# Effort Parameter Guidance (dino v18.2.0)
-
-## Overview
-
-The `effort` parameter guides Claude Code on how much thinking/reasoning to apply to a task. This maps to model selection and extended thinking token allocation.
-
-## Effort Levels
-
-| Level | Model | Thinking Tokens | Use Cases |
-|-------|-------|-----------------|-----------|
-| **low** | haiku | <2K tokens | Status checks, simple lookups, archival |
-| **medium** | sonnet | 2-5K tokens | Research, implementation, testing |
-| **high** | opus | 5K+ tokens | Planning, architecture, complex analysis |
-
-## When to Use Each Level
-
-### Low Effort
-- Quick status checks (`/dino.status`)
-- Simple lookups and reads
-- Archival tasks (`/dino.fossil`)
-- Focus changes (`/dino.status --focus`)
-- Clarification questions (`/dino.sniff`)
-- Loading handoff context (`/dino.handoff`)
-- Next step suggestions (`/dino.status --next`)
-- Context optimization guidance (`/dino.status --tips`)
-
-**Characteristics:**
-- Single-step operations
-- Minimal reasoning required
-- Speed is more important than depth
-- Read-heavy, write-light
-
-### Medium Effort
-- Research tasks (`/dino.scout`, `/research`)
-- Implementation execution (`/dino.hunt`)
-- Test verification (`/dino.aging`)
-- Task breakdown (`/dino.hatch`)
-- Release workflows (`/dino.release`)
-
-**Characteristics:**
-- Multi-step operations
-- Moderate reasoning required
-- Balance of speed and quality
-- Mix of reading and writing
-
-### High Effort
-- Architecture planning (`/dino.nest`)
-- Complex analysis
-- Design decisions
-- Multi-component changes
-
-**Characteristics:**
-- Deep reasoning required
-- Many interdependent decisions
-- Quality is more important than speed
-- Significant impact on codebase
-
-## Opus 4.5 Extended Thinking
-
-For high-effort tasks, Opus 4.5's extended thinking capabilities can be leveraged:
-
-```yaml
-# In SKILL.md frontmatter
-effort: high  # Signals to use extended thinking
-```
-
-**Extended thinking is useful for:**
-- Understanding complex requirements
-- Exploring multiple approaches
-- Identifying edge cases
-- Planning multi-file changes
-- Architectural decisions
-
-## Mapping to SKILL.md
-
-Each skill specifies its effort level in frontmatter:
-
-```yaml
----
-description: Plan implementation with structured approach
-model: opus
-effort: high
-context: fork
----
-```
-
-## Budget Considerations
-
-Effort level affects token budget:
-
-| Level | Typical Budget Impact | Cost Factor |
-|-------|----------------------|-------------|
-| low | ~1K tokens | 1x (cheapest) |
-| medium | ~5K tokens | 3x |
-| high | ~15K+ tokens | 10x |
-
-Use the lowest effort level that produces good results for the task.
-
-## Dynamic Effort Adjustment
-
-Consider increasing effort when:
-- Initial attempt fails or produces poor results
-- Task complexity is higher than expected
-- User feedback indicates need for deeper analysis
-
-Consider decreasing effort when:
-- Task is simpler than expected
-- Budget constraints are tight
-- Speed is prioritized over depth
+# Effort Parameter Guidance (dino v18.2.0)
+
+## Overview
+
+The `effort` parameter guides Claude Code on how much thinking/reasoning to apply to a task. This maps to model selection and extended thinking token allocation.
+
+## Effort Levels
+
+| Level | Model | Thinking Tokens | Use Cases |
+|-------|-------|-----------------|-----------|
+| **low** | haiku | <2K tokens | Status checks, simple lookups, archival |
+| **medium** | sonnet | 2-5K tokens | Research, implementation, testing |
+| **high** | opus | 5K+ tokens | Planning, architecture, complex analysis |
+
+## When to Use Each Level
+
+### Low Effort
+- Quick status checks (`/dino.status`)
+- Simple lookups and reads
+- Archival tasks (`/dino.fossil`)
+- Focus changes (`/dino.status --focus`)
+- Clarification questions (`/dino.sniff`)
+- Loading handoff context (`/dino.handoff`)
+- Next step suggestions (`/dino.status --next`)
+- Context optimization guidance (`/dino.status --tips`)
+
+**Characteristics:**
+- Single-step operations
+- Minimal reasoning required
+- Speed is more important than depth
+- Read-heavy, write-light
+
+### Medium Effort
+- Research tasks (`/dino.scout`, `/research`)
+- Implementation execution (`/dino.hunt`)
+- Test verification (`/dino.aging`)
+- Task breakdown (`/dino.hatch`)
+- Release workflows (`/dino.release`)
+
+**Characteristics:**
+- Multi-step operations
+- Moderate reasoning required
+- Balance of speed and quality
+- Mix of reading and writing
+
+### High Effort
+- Architecture planning (`/dino.nest`)
+- Complex analysis
+- Design decisions
+- Multi-component changes
+
+**Characteristics:**
+- Deep reasoning required
+- Many interdependent decisions
+- Quality is more important than speed
+- Significant impact on codebase
+
+## Opus 4.5 Extended Thinking
+
+For high-effort tasks, Opus 4.5's extended thinking capabilities can be leveraged:
+
+```yaml
+# In SKILL.md frontmatter
+effort: high  # Signals to use extended thinking
+```
+
+**Extended thinking is useful for:**
+- Understanding complex requirements
+- Exploring multiple approaches
+- Identifying edge cases
+- Planning multi-file changes
+- Architectural decisions
+
+## Mapping to SKILL.md
+
+Each skill specifies its effort level in frontmatter:
+
+```yaml
+---
+description: Plan implementation with structured approach
+model: opus
+effort: high
+context: fork
+---
+```
+
+## Budget Considerations
+
+Effort level affects token budget:
+
+| Level | Typical Budget Impact | Cost Factor |
+|-------|----------------------|-------------|
+| low | ~1K tokens | 1x (cheapest) |
+| medium | ~5K tokens | 3x |
+| high | ~15K+ tokens | 10x |
+
+Use the lowest effort level that produces good results for the task.
+
+## Dynamic Effort Adjustment
+
+Consider increasing effort when:
+- Initial attempt fails or produces poor results
+- Task complexity is higher than expected
+- User feedback indicates need for deeper analysis
+
+Consider decreasing effort when:
+- Task is simpler than expected
+- Budget constraints are tight
+- Speed is prioritized over depth

package/.claude/rules/dino/ralph-enforcement.md

@@ -1,77 +1,77 @@
-# Ralph Loop Enforcement (dino v18.2.0)
-
-## CRITICAL: Mandatory Ralph Gate
-
-When a user requests Ralph Loop or any complex multi-story task, you MUST use the `/dino.ralph-gate` skill. This is NOT optional.
-
-## Trigger Keywords (Case Insensitive)
-
-IMMEDIATELY invoke `/dino.ralph-gate` when user message contains:
-
-| Keyword | Example |
-|---------|---------|
-| "ralph loop" | "Start ralph loop for this feature" |
-| "ralph" | "Use ralph for this" |
-| "/dino.ralph" | "/dino.ralph" |
-| "fresh context" + "iteration" | "Use fresh context iterations" |
-| "multi-story" | "This is multi-story work" |
-| "three agent" | "Use three agent system" |
-| "gated implementation" | "I want gated implementation" |
-
-## Complexity Auto-Trigger
-
-ALSO invoke `/dino.ralph-gate` when task analysis shows:
-- Complexity score > 70
-- File count > 10
-- Distinct parts >= 3
-- Breaking changes + scope >= 15
-
-## Forbidden Bypass
-
-You MUST NOT:
-- Skip directly to `/dino.ralph` without gating
-- Start implementation without auditor approval
-- Proceed with ambiguity score > 0
-- Accept confidence < 95% for Ralph-eligible tasks
-
-## Three-Agent Requirement
-
-ALL THREE agents must execute in sequence:
-
-```
-1. SCOUT (Research)     -> Deep exploration, pattern mining
-       |
-2. AUDITOR (Gatekeeper) -> 95% score, zero ambiguity required
-       |
-3. RALPH (Executor)     -> Only after GATE: APPROVED
-```
-
-If auditor rejects, loop back to scout/clarification. Max 3 cycles.
-
-## Skill Invocation
-
-When trigger detected, use Skill tool:
-
-```
-Skill: dino.ralph-gate
-```
-
-Do NOT use regular `/dino.ralph` for complex tasks - always use `/dino.ralph-gate`.
-
-## Exception: Simple Ralph
-
-For tasks that:
-- Already have complete PRD with stories
-- Already went through spec with 90%+ confidence
-- User explicitly says "skip gate" or "already specified"
-
-You MAY use `/dino.ralph` directly, but MUST confirm with user first.
-
-## Audit Trail
-
-All Ralph Gate executions produce:
-- `.dino/ralph-gate/research-report.md`
-- `.dino/ralph-gate/audit-report.md`
-- `.dino/ralph-gate/validated-prd.md`
-
-These files provide accountability and can be reviewed.
+# Ralph Loop Enforcement (dino v18.2.0)
+
+## CRITICAL: Mandatory Ralph Gate
+
+When a user requests Ralph Loop or any complex multi-story task, you MUST use the `/dino.ralph-gate` skill. This is NOT optional.
+
+## Trigger Keywords (Case Insensitive)
+
+IMMEDIATELY invoke `/dino.ralph-gate` when user message contains:
+
+| Keyword | Example |
+|---------|---------|
+| "ralph loop" | "Start ralph loop for this feature" |
+| "ralph" | "Use ralph for this" |
+| "/dino.ralph" | "/dino.ralph" |
+| "fresh context" + "iteration" | "Use fresh context iterations" |
+| "multi-story" | "This is multi-story work" |
+| "three agent" | "Use three agent system" |
+| "gated implementation" | "I want gated implementation" |
+
+## Complexity Auto-Trigger
+
+ALSO invoke `/dino.ralph-gate` when task analysis shows:
+- Complexity score > 70
+- File count > 10
+- Distinct parts >= 3
+- Breaking changes + scope >= 15
+
+## Forbidden Bypass
+
+You MUST NOT:
+- Skip directly to `/dino.ralph` without gating
+- Start implementation without auditor approval
+- Proceed with ambiguity score > 0
+- Accept confidence < 95% for Ralph-eligible tasks
+
+## Three-Agent Requirement
+
+ALL THREE agents must execute in sequence:
+
+```
+1. SCOUT (Research)     -> Deep exploration, pattern mining
+       |
+2. AUDITOR (Gatekeeper) -> 95% score, zero ambiguity required
+       |
+3. RALPH (Executor)     -> Only after GATE: APPROVED
+```
+
+If auditor rejects, loop back to scout/clarification. Max 3 cycles.
+
+## Skill Invocation
+
+When trigger detected, use Skill tool:
+
+```
+Skill: dino.ralph-gate
+```
+
+Do NOT use regular `/dino.ralph` for complex tasks - always use `/dino.ralph-gate`.
+
+## Exception: Simple Ralph
+
+For tasks that:
+- Already have complete PRD with stories
+- Already went through spec with 90%+ confidence
+- User explicitly says "skip gate" or "already specified"
+
+You MAY use `/dino.ralph` directly, but MUST confirm with user first.
+
+## Audit Trail
+
+All Ralph Gate executions produce:
+- `.dino/ralph-gate/research-report.md`
+- `.dino/ralph-gate/audit-report.md`
+- `.dino/ralph-gate/validated-prd.md`
+
+These files provide accountability and can be reviewed.