compact-agent 1.10.1 → 1.11.0

package/resources/ecc/commands/learn-eval.md

@@ -0,0 +1,116 @@
+---
+description: "Extract reusable patterns from the session, self-evaluate quality before saving, and determine the right save location (Global vs Project)."
+---
+
+# /learn-eval - Extract, Evaluate, then Save
+
+Extends `/learn` with a quality gate, save-location decision, and knowledge-placement awareness before writing any skill file.
+
+## What to Extract
+
+Look for:
+
+1. **Error Resolution Patterns** — root cause + fix + reusability
+2. **Debugging Techniques** — non-obvious steps, tool combinations
+3. **Workarounds** — library quirks, API limitations, version-specific fixes
+4. **Project-Specific Patterns** — conventions, architecture decisions, integration patterns
+
+## Process
+
+1. Review the session for extractable patterns
+2. Identify the most valuable/reusable insight
+
+3. **Determine save location:**
+   - Ask: "Would this pattern be useful in a different project?"
+   - **Global** (`~/.claude/skills/learned/`): Generic patterns usable across 2+ projects (bash compatibility, LLM API behavior, debugging techniques, etc.)
+   - **Project** (`.claude/skills/learned/` in current project): Project-specific knowledge (quirks of a particular config file, project-specific architecture decisions, etc.)
+   - When in doubt, choose Global (moving Global → Project is easier than the reverse)
+
+4. Draft the skill file using this format:
+
+```markdown
+---
+name: pattern-name
+description: "Under 130 characters"
+user-invocable: false
+origin: auto-extracted
+---
+
+# [Descriptive Pattern Name]
+
+**Extracted:** [Date]
+**Context:** [Brief description of when this applies]
+
+## Problem
+[What problem this solves - be specific]
+
+## Solution
+[The pattern/technique/workaround - with code examples]
+
+## When to Use
+[Trigger conditions]
+```
+
+5. **Quality gate — Checklist + Holistic verdict**
+
+   ### 5a. Required checklist (verify by actually reading files)
+
+   Execute **all** of the following before evaluating the draft:
+
+   - [ ] Grep `~/.claude/skills/` and relevant project `.claude/skills/` files by keyword to check for content overlap
+   - [ ] Check MEMORY.md (both project and global) for overlap
+   - [ ] Consider whether appending to an existing skill would suffice
+   - [ ] Confirm this is a reusable pattern, not a one-off fix
+
+   ### 5b. Holistic verdict
+
+   Synthesize the checklist results and draft quality, then choose **one** of the following:
+
+   | Verdict | Meaning | Next Action |
+   |---------|---------|-------------|
+   | **Save** | Unique, specific, well-scoped | Proceed to Step 6 |
+   | **Improve then Save** | Valuable but needs refinement | List improvements → revise → re-evaluate (once) |
+   | **Absorb into [X]** | Should be appended to an existing skill | Show target skill and additions → Step 6 |
+   | **Drop** | Trivial, redundant, or too abstract | Explain reasoning and stop |
+
+**Guideline dimensions** (informing the verdict, not scored):
+
+- **Specificity & Actionability**: Contains code examples or commands that are immediately usable
+- **Scope Fit**: Name, trigger conditions, and content are aligned and focused on a single pattern
+- **Uniqueness**: Provides value not covered by existing skills (informed by checklist results)
+- **Reusability**: Realistic trigger scenarios exist in future sessions
+
+6. **Verdict-specific confirmation flow**
+
+- **Improve then Save**: Present the required improvements + revised draft + updated checklist/verdict after one re-evaluation; if the revised verdict is **Save**, save after user confirmation, otherwise follow the new verdict
+- **Save**: Present save path + checklist results + 1-line verdict rationale + full draft → save after user confirmation
+- **Absorb into [X]**: Present target path + additions (diff format) + checklist results + verdict rationale → append after user confirmation
+- **Drop**: Show checklist results + reasoning only (no confirmation needed)
+
+7. Save / Absorb to the determined location
+
+## Output Format for Step 5
+
+```
+### Checklist
+- [x] skills/ grep: no overlap (or: overlap found → details)
+- [x] MEMORY.md: no overlap (or: overlap found → details)
+- [x] Existing skill append: new file appropriate (or: should append to [X])
+- [x] Reusability: confirmed (or: one-off → Drop)
+
+### Verdict: Save / Improve then Save / Absorb into [X] / Drop
+
+**Rationale:** (1-2 sentences explaining the verdict)
+```
+
+## Design Rationale
+
+This version replaces the previous 5-dimension numeric scoring rubric (Specificity, Actionability, Scope Fit, Non-redundancy, Coverage scored 1-5) with a checklist-based holistic verdict system. Modern frontier models (Opus 4.6+) have strong contextual judgment — forcing rich qualitative signals into numeric scores loses nuance and can produce misleading totals. The holistic approach lets the model weigh all factors naturally, producing more accurate save/drop decisions while the explicit checklist ensures no critical check is skipped.
+
+## Notes
+
+- Don't extract trivial fixes (typos, simple syntax errors)
+- Don't extract one-time issues (specific API outages, etc.)
+- Focus on patterns that will save time in future sessions
+- Keep skills focused — one pattern per skill
+- When the verdict is Absorb, append to the existing skill rather than creating a new file

package/resources/ecc/commands/learn.md

@@ -0,0 +1,74 @@
+---
+description: Extract reusable patterns from the current session and save them as candidate skills or guidance.
+---
+
+# /learn - Extract Reusable Patterns
+
+Analyze the current session and extract any patterns worth saving as skills.
+
+## Trigger
+
+Run `/learn` at any point during a session when you've solved a non-trivial problem.
+
+## What to Extract
+
+Look for:
+
+1. **Error Resolution Patterns**
+   - What error occurred?
+   - What was the root cause?
+   - What fixed it?
+   - Is this reusable for similar errors?
+
+2. **Debugging Techniques**
+   - Non-obvious debugging steps
+   - Tool combinations that worked
+   - Diagnostic patterns
+
+3. **Workarounds**
+   - Library quirks
+   - API limitations
+   - Version-specific fixes
+
+4. **Project-Specific Patterns**
+   - Codebase conventions discovered
+   - Architecture decisions made
+   - Integration patterns
+
+## Output Format
+
+Create a skill file at `~/.claude/skills/learned/[pattern-name].md`:
+
+```markdown
+# [Descriptive Pattern Name]
+
+**Extracted:** [Date]
+**Context:** [Brief description of when this applies]
+
+## Problem
+[What problem this solves - be specific]
+
+## Solution
+[The pattern/technique/workaround]
+
+## Example
+[Code example if applicable]
+
+## When to Use
+[Trigger conditions - what should activate this skill]
+```
+
+## Process
+
+1. Review the session for extractable patterns
+2. Identify the most valuable/reusable insight
+3. Draft the skill file
+4. Ask user to confirm before saving
+5. Save to `~/.claude/skills/learned/`
+
+## Notes
+
+- Don't extract trivial fixes (typos, simple syntax errors)
+- Don't extract one-time issues (specific API outages, etc.)
+- Focus on patterns that will save time in future sessions
+- Keep skills focused - one pattern per skill

package/resources/ecc/commands/loop-start.md

@@ -0,0 +1,36 @@
+---
+description: Start a managed autonomous loop pattern with safety defaults and explicit stop conditions.
+---
+
+# Loop Start Command
+
+Start a managed autonomous loop pattern with safety defaults.
+
+## Usage
+
+`/loop-start [pattern] [--mode safe|fast]`
+
+- `pattern`: `sequential`, `continuous-pr`, `rfc-dag`, `infinite`
+- `--mode`:
+  - `safe` (default): strict quality gates and checkpoints
+  - `fast`: reduced gates for speed
+
+## Flow
+
+1. Confirm repository state and branch strategy.
+2. Select loop pattern and model tier strategy.
+3. Enable required hooks/profile for the chosen mode.
+4. Create loop plan and write runbook under `.claude/plans/`.
+5. Print commands to start and monitor the loop.
+
+## Required Safety Checks
+
+- Verify tests pass before first loop iteration.
+- Ensure `ECC_HOOK_PROFILE` is not disabled globally.
+- Ensure loop has explicit stop condition.
+
+## Arguments
+
+$ARGUMENTS:
+- `<pattern>` optional (`sequential|continuous-pr|rfc-dag|infinite`)
+- `--mode safe|fast` optional

package/resources/ecc/commands/loop-status.md

@@ -0,0 +1,77 @@
+---
+description: Inspect active loop state, progress, failure signals, and recommended intervention.
+---
+
+# Loop Status Command
+
+Inspect active loop state, progress, and failure signals.
+
+This slash command can only run after the current session dequeues it. If you
+need to inspect a wedged or sibling session, run the packaged CLI from another
+terminal:
+
+```bash
+npx --package ecc-universal ecc loop-status --json
+```
+
+The CLI scans local Claude transcript JSONL files under
+`~/.claude/projects/**` and reports stale `ScheduleWakeup` calls or `Bash`
+tool calls that have no matching `tool_result`.
+
+## Usage
+
+`/loop-status [--watch]`
+
+## What to Report
+
+- active loop pattern
+- current phase and last successful checkpoint
+- failing checks (if any)
+- estimated time/cost drift
+- recommended intervention (continue/pause/stop)
+
+## Cross-Session CLI
+
+- `ecc loop-status --json` emits machine-readable status for recent local
+  Claude transcripts.
+- `ecc loop-status --home <dir>` scans a different home directory when
+  inspecting another local profile or mounted workspace.
+- `ecc loop-status --transcript <session.jsonl>` inspects one transcript
+  directly.
+- `ecc loop-status --bash-timeout-seconds 1800` adjusts the stale Bash
+  threshold.
+- `ecc loop-status --exit-code` exits `2` when stale loop or tool signals are
+  found, or `1` when transcripts cannot be scanned.
+- `--exit-code` with `--watch` requires `--watch-count` so watchdog scripts do
+  not wait forever for a process exit.
+- `ecc loop-status --watch` refreshes status until interrupted.
+- `ecc loop-status --watch --watch-count 3 --exit-code` refreshes a bounded
+  number of times, then exits with the highest status seen.
+- `ecc loop-status --watch --watch-count 3` emits a bounded watch stream for
+  scripts and handoffs.
+- `ecc loop-status --watch --write-dir ~/.claude/loops` maintains
+  `index.json` and per-session JSON snapshots for sibling terminals or
+  watchdog scripts.
+
+## Watch Mode
+
+When `--watch` is present, refresh status periodically. With `--json`, each
+refresh is emitted as one JSON object per line so another terminal or script can
+consume the stream.
+
+## Snapshot Files
+
+Use `--write-dir <dir>` when a separate process needs to inspect loop state
+without waiting for the current Claude session to dequeue `/loop-status`. The
+CLI writes:
+
+- `index.json` with one row per inspected session.
+- `<session-id>.json` with the full status payload for that session.
+
+These files are snapshots of local transcript analysis. They do not control or
+timeout Claude Code runtime tool calls.
+
+## Arguments
+
+$ARGUMENTS:
+- `--watch` optional

package/resources/ecc/commands/model-route.md

@@ -0,0 +1,30 @@
+---
+description: Recommend the best model tier for the current task based on complexity, risk, and budget.
+---
+
+# Model Route Command
+
+Recommend the best model tier for the current task by complexity and budget.
+
+## Usage
+
+`/model-route [task-description] [--budget low|med|high]`
+
+## Routing Heuristic
+
+- `haiku`: deterministic, low-risk mechanical changes
+- `sonnet`: default for implementation and refactors
+- `opus`: architecture, deep review, ambiguous requirements
+
+## Required Output
+
+- recommended model
+- confidence level
+- why this model fits
+- fallback model if first attempt fails
+
+## Arguments
+
+$ARGUMENTS:
+- `[task-description]` optional free-text
+- `--budget low|med|high` optional

package/resources/ecc/commands/multi-backend.md

@@ -0,0 +1,162 @@
+---
+description: Run a backend-focused multi-model workflow for APIs, algorithms, data, and business logic.
+---
+
+# Backend - Backend-Focused Development
+
+Backend-focused workflow (Research → Ideation → Plan → Execute → Optimize → Review), Codex-led.
+
+## Usage
+
+```bash
+/backend <backend task description>
+```
+
+## Context
+
+- Backend task: $ARGUMENTS
+- Codex-led, Gemini for auxiliary reference
+- Applicable: API design, algorithm implementation, database optimization, business logic
+
+## Your Role
+
+You are the **Backend Orchestrator**, coordinating multi-model collaboration for server-side tasks (Research → Ideation → Plan → Execute → Optimize → Review).
+
+**Collaborative Models**:
+- **Codex** – Backend logic, algorithms (**Backend authority, trustworthy**)
+- **Gemini** – Frontend perspective (**Backend opinions for reference only**)
+- **Claude (self)** – Orchestration, planning, execution, delivery
+
+---
+
+## Multi-Model Call Specification
+
+**Call Syntax**:
+
+```
+# New session call
+Bash({
+  command: "~/.claude/bin/codeagent-wrapper {{LITE_MODE_FLAG}}--backend codex - \"$PWD\" <<'EOF'
+ROLE_FILE: <role prompt path>
+<TASK>
+Requirement: <enhanced requirement (or $ARGUMENTS if not enhanced)>
+Context: <project context and analysis from previous phases>
+</TASK>
+OUTPUT: Expected output format
+EOF",
+  run_in_background: false,
+  timeout: 3600000,
+  description: "Brief description"
+})
+
+# Resume session call
+Bash({
+  command: "~/.claude/bin/codeagent-wrapper {{LITE_MODE_FLAG}}--backend codex resume <SESSION_ID> - \"$PWD\" <<'EOF'
+ROLE_FILE: <role prompt path>
+<TASK>
+Requirement: <enhanced requirement (or $ARGUMENTS if not enhanced)>
+Context: <project context and analysis from previous phases>
+</TASK>
+OUTPUT: Expected output format
+EOF",
+  run_in_background: false,
+  timeout: 3600000,
+  description: "Brief description"
+})
+```
+
+**Role Prompts**:
+
+| Phase | Codex |
+|-------|-------|
+| Analysis | `~/.claude/.ccg/prompts/codex/analyzer.md` |
+| Planning | `~/.claude/.ccg/prompts/codex/architect.md` |
+| Review | `~/.claude/.ccg/prompts/codex/reviewer.md` |
+
+**Session Reuse**: Each call returns `SESSION_ID: xxx`, use `resume xxx` for subsequent phases. Save `CODEX_SESSION` in Phase 2, use `resume` in Phases 3 and 5.
+
+---
+
+## Communication Guidelines
+
+1. Start responses with mode label `[Mode: X]`, initial is `[Mode: Research]`
+2. Follow strict sequence: `Research → Ideation → Plan → Execute → Optimize → Review`
+3. Use `AskUserQuestion` tool for user interaction when needed (e.g., confirmation/selection/approval)
+
+---
+
+## Core Workflow
+
+### Phase 0: Prompt Enhancement (Optional)
+
+`[Mode: Prepare]` - If ace-tool MCP available, call `mcp__ace-tool__enhance_prompt`, **replace original $ARGUMENTS with enhanced result for subsequent Codex calls**. If unavailable, use `$ARGUMENTS` as-is.
+
+### Phase 1: Research
+
+`[Mode: Research]` - Understand requirements and gather context
+
+1. **Code Retrieval** (if ace-tool MCP available): Call `mcp__ace-tool__search_context` to retrieve existing APIs, data models, service architecture. If unavailable, use built-in tools: `Glob` for file discovery, `Grep` for symbol/API search, `Read` for context gathering, `Task` (Explore agent) for deeper exploration.
+2. Requirement completeness score (0-10): >=7 continue, <7 stop and supplement
+
+### Phase 2: Ideation
+
+`[Mode: Ideation]` - Codex-led analysis
+
+**MUST call Codex** (follow call specification above):
+- ROLE_FILE: `~/.claude/.ccg/prompts/codex/analyzer.md`
+- Requirement: Enhanced requirement (or $ARGUMENTS if not enhanced)
+- Context: Project context from Phase 1
+- OUTPUT: Technical feasibility analysis, recommended solutions (at least 2), risk assessment
+
+**Save SESSION_ID** (`CODEX_SESSION`) for subsequent phase reuse.
+
+Output solutions (at least 2), wait for user selection.
+
+### Phase 3: Planning
+
+`[Mode: Plan]` - Codex-led planning
+
+**MUST call Codex** (use `resume <CODEX_SESSION>` to reuse session):
+- ROLE_FILE: `~/.claude/.ccg/prompts/codex/architect.md`
+- Requirement: User's selected solution
+- Context: Analysis results from Phase 2
+- OUTPUT: File structure, function/class design, dependency relationships
+
+Claude synthesizes plan, save to `.claude/plan/task-name.md` after user approval.
+
+### Phase 4: Implementation
+
+`[Mode: Execute]` - Code development
+
+- Strictly follow approved plan
+- Follow existing project code standards
+- Ensure error handling, security, performance optimization
+
+### Phase 5: Optimization
+
+`[Mode: Optimize]` - Codex-led review
+
+**MUST call Codex** (follow call specification above):
+- ROLE_FILE: `~/.claude/.ccg/prompts/codex/reviewer.md`
+- Requirement: Review the following backend code changes
+- Context: git diff or code content
+- OUTPUT: Security, performance, error handling, API compliance issues list
+
+Integrate review feedback, execute optimization after user confirmation.
+
+### Phase 6: Quality Review
+
+`[Mode: Review]` - Final evaluation
+
+- Check completion against plan
+- Run tests to verify functionality
+- Report issues and recommendations
+
+---
+
+## Key Rules
+
+1. **Codex backend opinions are trustworthy**
+2. **Gemini backend opinions for reference only**
+3. External models have **zero filesystem write access**
+4. Claude handles all code writes and file operations