claude-turing 4.5.0 → 4.7.0

package/skills/turing/export/SKILL.md

@@ -0,0 +1,48 @@
+---
+name: export
+description: Export model to production format with equivalence verification, latency benchmarking, and deployment model card.
+disable-model-invocation: true
+argument-hint: "[exp-id] [--format joblib|xgboost_json|onnx|torchscript|tflite]"
+allowed-tools: Read, Bash(*), Grep, Glob
+---
+
+Export a trained model to a production-ready format.
+
+## Steps
+
+1. **Activate environment:**
+   ```bash
+   source .venv/bin/activate
+   ```
+
+2. **Parse arguments from `$ARGUMENTS`:**
+   - First argument can be an experiment ID (e.g., `exp-042`); defaults to best
+   - `--format joblib|xgboost_json|onnx|torchscript|tflite` specifies export format (auto-detected if omitted)
+   - `--skip-equivalence` skips inference equivalence check
+   - `--skip-latency` skips latency benchmark
+   - `--samples 100` sets test sample count
+
+3. **Run export pipeline:**
+   ```bash
+   python scripts/export_model.py $ARGUMENTS
+   ```
+
+4. **Report results:**
+   - **Export:** format, file size, output path, dependencies
+   - **Equivalence:** verdict (equivalent/approximately_equivalent/divergent), max delta
+   - **Latency:** p50/p95/p99 ms, speedup vs original
+   - **Model Card:** metrics, seed study, equivalence, latency, dependencies
+
+5. **Output:** exported model + model_card.yaml written to `exports/exp-NNN/`
+
+6. **If model file not found:** suggest checking models/best/ directory.
+
+## Examples
+
+```
+/turing:export                                     # Best experiment, default format
+/turing:export exp-042                             # Specific experiment
+/turing:export --format xgboost_json               # Native XGBoost JSON
+/turing:export --format onnx                       # ONNX format
+/turing:export --skip-equivalence --skip-latency   # Fast export
+```

package/skills/turing/feature/SKILL.md

@@ -0,0 +1,42 @@
+---
+name: feature
+description: Automated feature selection — multi-method importance consensus, redundancy detection, and interaction feature generation.
+disable-model-invocation: true
+argument-hint: "[--method all|importance] [--top-k 20]"
+allowed-tools: Read, Bash(*), Grep, Glob
+---
+
+Systematically evaluate which features matter and which are noise.
+
+## Steps
+
+1. **Activate environment:**
+   ```bash
+   source .venv/bin/activate
+   ```
+
+2. **Parse arguments from `$ARGUMENTS`:**
+   - `--method all|importance|selection|generation` — analysis type (default: all)
+   - `--top-k 20` — number of top features to consider
+   - `--json` — raw JSON output
+
+3. **Run feature analysis:**
+   ```bash
+   python scripts/feature_intelligence.py $ARGUMENTS
+   ```
+
+4. **Report includes:**
+   - Consensus ranking: features ranked by number of methods placing them in top-K
+   - Per-method ranks: mutual information, L1, tree-based
+   - Redundant pairs: features with |r| > 0.95
+   - Candidate interaction features from top consensus set
+   - Drop recommendation for zero-consensus features
+
+5. **Saved output:** report in `experiments/features/features-*.yaml`
+
+## Examples
+
+```
+/turing:feature                      # Full analysis
+/turing:feature --top-k 10           # Top-10 consensus
+```

package/skills/turing/flashback/SKILL.md

@@ -0,0 +1,22 @@
+---
+name: flashback
+description: Session context restoration — "where was I?" summary after days away. Current best, pending hypotheses, last session, annotations.
+disable-model-invocation: true
+argument-hint: "[--days 7] [--last 10]"
+allowed-tools: Read, Bash(*), Grep, Glob
+---
+
+Come back to a project after a week and start working in 10 seconds instead of 30 minutes.
+
+## Steps
+1. **Activate environment:** `source .venv/bin/activate`
+2. **Run:** `python scripts/session_flashback.py $ARGUMENTS`
+3. **Report:** current best, last session experiments, pending hypotheses, annotations, budget, suggested next action
+4. **Saved output:** `experiments/flashbacks/flashback-*.yaml`
+
+## Examples
+```
+/turing:flashback                    # Default: last 7 days
+/turing:flashback --days 14          # 2-week lookback
+/turing:flashback --last 5           # Last 5 experiments
+```

package/skills/turing/fork/SKILL.md

@@ -0,0 +1,40 @@
+---
+name: fork
+description: Branch an experiment into parallel tracks — run both A and B, report the winner.
+disable-model-invocation: true
+argument-hint: "<exp-id> --branches \"approach A\" \"approach B\" [--auto-promote]"
+allowed-tools: Read, Bash(*), Grep, Glob
+---
+
+Fork an experiment into parallel branches and compare results.
+
+## Steps
+
+1. **Activate environment:**
+   ```bash
+   source .venv/bin/activate
+   ```
+
+2. **Parse arguments from `$ARGUMENTS`:**
+   - First argument is the parent experiment ID
+   - `--branches "A" "B" "C"` — branch descriptions (2+ required)
+   - `--auto-promote` — automatically keep the winning branch
+
+3. **Run fork:**
+   ```bash
+   python scripts/fork_experiment.py $ARGUMENTS
+   ```
+
+4. **Report results:**
+   - Comparison tree showing each branch's metric
+   - Winner identified and marked
+   - Recommendation: promote winner, abandon rest
+
+5. **Saved output:** report written to `experiments/forks/exp-NNN-fork.yaml`
+
+## Examples
+
+```
+/turing:fork exp-042 --branches "LightGBM with dart" "XGBoost deeper trees"
+/turing:fork exp-042 --branches "A" "B" "C" --auto-promote
+```

package/skills/turing/frontier/SKILL.md

@@ -0,0 +1,45 @@
+---
+name: frontier
+description: Visualize Pareto frontier across multiple objectives — answers "which model is actually best?" when there are tradeoffs.
+disable-model-invocation: true
+argument-hint: "[--metrics \"accuracy,train_seconds,n_params\"] [--ascii]"
+allowed-tools: Read, Bash(*), Grep, Glob
+---
+
+Visualize the Pareto frontier across multiple objectives from experiment history.
+
+## Steps
+
+1. **Activate environment:**
+   ```bash
+   source .venv/bin/activate
+   ```
+
+2. **Parse arguments from `$ARGUMENTS`:**
+   - `--metrics "accuracy,train_seconds,n_params"` specifies metrics to analyze
+   - Without `--metrics`, uses primary metric + train_seconds from config
+   - `--ascii` generates an ASCII scatter plot (2D projection)
+
+3. **Run Pareto analysis:**
+   ```bash
+   python scripts/pareto_frontier.py $ARGUMENTS
+   ```
+
+4. **Report results:**
+   - **Pareto-optimal experiments:** table with all metrics and what each is best at
+   - **Dominated experiments:** with their nearest Pareto neighbor
+   - **ASCII scatter plot** (if `--ascii`): 2D projection with * for Pareto, · for dominated
+   - Summary: "N Pareto-optimal of M experiments across K metrics"
+
+5. **Saved output:** results written to `experiments/frontiers/frontier-YYYY-MM-DD.yaml`
+
+6. **If no experiments have all requested metrics:** suggest which metrics are available.
+
+## Examples
+
+```
+/turing:frontier                                              # Default: metric vs time
+/turing:frontier --metrics "accuracy,train_seconds"           # 2D frontier
+/turing:frontier --metrics "accuracy,train_seconds,n_params"  # 3D frontier
+/turing:frontier --ascii                                      # With scatter plot
+```

package/skills/turing/init/SKILL.md

@@ -0,0 +1,154 @@
+---
+name: init
+description: Initialize a new ML project with the Turing autoresearch harness. Scaffolds the full experiment infrastructure — immutable evaluation pipeline, agent-editable training code, structured logging, convergence detection hooks, and a Python virtual environment. Use --plan to generate a research plan.
+disable-model-invocation: true
+argument-hint: "[project_name] [--plan]"
+allowed-tools: Read, Write, Edit, Bash(*), Grep, Glob, WebSearch, WebFetch
+---
+
+Scaffold a new ML project with the Turing autoresearch harness. This creates the separation between the measurement apparatus (READ-ONLY) and the hypothesis space (AGENT-EDITABLE) that makes autonomous experimentation trustworthy.
+
+## Interactive Setup
+
+Ask the user for the following (or accept from `$ARGUMENTS` if provided as JSON):
+
+1. **Project name** (`{{PROJECT_NAME}}`): Name of the ML project (e.g., "sentiment", "churn", "fraud-detection")
+2. **Target metric** (`{{TARGET_METRIC}}`): Primary metric to optimize (e.g., "accuracy", "f1", "mae", "mse", "auc")
+3. **Metric direction**: Is lower better (mae, mse, loss) or higher better (accuracy, f1, auc)?
+4. **Task description** (`{{TASK_DESCRIPTION}}`): What the model does (e.g., "Predict customer churn from usage data")
+5. **ML directory** (`{{ML_DIR}}`): Where ML files go relative to project root (e.g., "ml/sentiment")
+6. **Data source** (`{{DATA_SOURCE}}`): Where training data comes from (e.g., "data/reviews.csv")
+
+## Scaffolding
+
+Once you have all 6 values, delegate to the unified scaffolding script:
+
+```bash
+python3 <templates_dir>/scripts/scaffold.py \
+    --project-name "<project_name>" \
+    --target-metric "<target_metric>" \
+    --metric-direction "<metric_direction>" \
+    --task-description "<task_description>" \
+    --ml-dir "<ml_dir>" \
+    --data-source "<data_source>" \
+    --templates-dir "<templates_dir>"
+```
+
+The scaffold script handles everything in a single atomic operation:
+- Copies all template files with placeholder substitution
+- Creates data/, experiments/, models/ directories
+- Sets up agent memory at `.claude/agent-memory/ml-researcher-{project_name}/MEMORY.md`
+- Configures Claude Code hooks in `.claude/settings.local.json`
+- Creates Python virtual environment and installs requirements
+- Verifies all placeholders were replaced (fails loudly if any remain)
+
+## Locating Templates
+
+Use the installed command-pack templates directory first:
+```
+.claude/commands/turing/templates/
+~/.claude/commands/turing/templates/
+```
+Then fall back to plugin or npm locations:
+```
+~/.claude/plugins/*/templates/
+node_modules/claude-turing/templates/
+```
+
+Example command:
+
+```bash
+python3 ~/.claude/commands/turing/templates/scripts/scaffold.py \
+    --project-name "<project_name>" \
+    --target-metric "<target_metric>" \
+    --metric-direction "<metric_direction>" \
+    --task-description "<task_description>" \
+    --ml-dir "<ml_dir>" \
+    --data-source "<data_source>" \
+    --templates-dir ~/.claude/commands/turing/templates
+```
+
+## After Scaffolding
+
+Report what was created:
+- The separation: READ-ONLY (`prepare.py`, `evaluate.py`) vs AGENT-EDITABLE (`train.py`)
+- Next steps: add data to the configured data source path, run `python prepare.py`, then `/turing:train`
+- The taste-leverage loop: `/turing:try` to inject hypotheses, `/turing:brief` for intelligence reports
+
+## Research Plan Generation (--plan flag)
+
+If `$ARGUMENTS` contains `--plan`, generate a research plan AFTER scaffolding. This gives the agent strategic direction for its first 5-10 experiments rather than ad-hoc exploration.
+
+### Steps:
+
+1. **Read the task context** from the just-created `config.yaml`: task description, model type, target metric, data source.
+
+2. **Search literature** with `WebSearch` for the task domain:
+   - "state of the art <task description> machine learning 2024 2025"
+   - "best model <target metric> <data type> benchmark"
+   - "<task description> common approaches survey"
+
+   Use `WebFetch` on top 2-3 results to extract: dominant model families, typical metric ranges, known challenges.
+
+3. **Generate `RESEARCH_PLAN.md`** in the ML project directory with this structure:
+
+   ```markdown
+   # Research Plan: <task description>
+
+   Generated: <date>
+
+   ## Task Summary
+   <one paragraph describing the task, data, and success criteria>
+
+   ## Model Families to Explore
+   Ordered by expected relevance based on literature:
+   1. **<family 1>** — <why, with citation>
+   2. **<family 2>** — <why, with citation>
+   3. **<family 3>** — <why, with citation>
+
+   ## Evaluation Strategy
+   - Primary metric: <metric> (<higher/lower> is better)
+   - Multi-run recommendation: <yes/no, based on expected variance>
+   - Baseline target: <realistic first-pass metric from literature>
+
+   ## Search Budget
+   - <N> experiments per model family before moving on
+   - Total budget: <N> experiments before first convergence check
+
+   ## Success Criteria
+   - Target metric: <value from literature benchmarks>
+   - Convergence: <patience> consecutive non-improvements
+
+   ## Known Challenges
+   - <challenge 1 from literature, e.g., "class imbalance common in this domain">
+   - <challenge 2>
+
+   ## Sources
+   - <citation 1>
+   - <citation 2>
+   ```
+
+4. **Self-critique the plan** (one round):
+   - Are the model families ordered by evidence strength?
+   - Is the budget realistic?
+   - Are the success criteria grounded in benchmark data?
+   Revise if any section is vague or unsupported.
+
+5. **Report:** "Research plan generated at `<ml_dir>/RESEARCH_PLAN.md`. The agent will read this during `/turing:train` for strategic direction."
+
+### Integration
+
+The agent's `program.md` OBSERVE step reads `RESEARCH_PLAN.md` (if it exists) for strategic direction. The plan is advisory — the agent can deviate but should note why in `experiment_state.yaml`.
+
+## Multiple Projects
+
+You can scaffold multiple ML projects in the same repository:
+
+```bash
+/turing:init    # First project: prompts for ml_dir (e.g., ml/sentiment)
+/turing:init    # Second project: prompts for ml_dir (e.g., ml/churn)
+```
+
+Each project gets its own directory with independent config, data, experiments, and models. `/turing:train ml/sentiment` or `/turing:train ml/churn` targets a specific project. If you `cd ml/sentiment` first, `/turing:train` auto-detects from cwd.
+
+Agent memory is scoped per project: `.claude/agent-memory/ml-researcher-{project_name}/MEMORY.md`

package/skills/turing/leak/SKILL.md

@@ -0,0 +1,47 @@
+---
+name: leak
+description: Targeted leakage detection — probe for data leakage with single-feature tests, correlation checks, and train/test overlap detection.
+disable-model-invocation: true
+argument-hint: "[--deep] [--features feature_1,feature_2]"
+allowed-tools: Read, Bash(*), Grep, Glob
+---
+
+Actively probe for data leakage. The #1 cause of "too good to be true" results.
+
+## Steps
+
+1. **Activate environment:**
+   ```bash
+   source .venv/bin/activate
+   ```
+
+2. **Parse arguments from `$ARGUMENTS`:**
+   - `--deep` — run full single-feature analysis (slow but thorough)
+   - `--features "feat_1,feat_2"` — check specific features
+   - `--json` — raw JSON output
+
+3. **Run leakage scan:**
+   ```bash
+   python scripts/leakage_detector.py $ARGUMENTS
+   ```
+
+4. **Checks performed:**
+   - **Feature-target correlation:** flag features with >0.95 correlation to target
+   - **Single-feature predictiveness (--deep):** train on each feature alone, flag any that achieve >80% of full model performance
+   - **Train/test overlap:** hash-based deduplication across splits
+
+5. **Verdicts:**
+   - **CLEAN** — no leakage detected
+   - **SUSPICIOUS** — warnings to review
+   - **LEAKAGE DETECTED** — critical flags found
+
+6. **Integration:** satisfies the "data leakage" check in `/turing:audit`
+
+7. **Saved output:** report in `experiments/leakage/leak-*.yaml`
+
+## Examples
+
+```
+/turing:leak                    # Standard correlation + overlap checks
+/turing:leak --deep             # Full single-feature analysis
+```

package/skills/turing/lit/SKILL.md

@@ -0,0 +1,47 @@
+---
+name: lit
+description: Literature search scoped to the current experiment domain — find papers, SOTA baselines, and related work without leaving the terminal.
+disable-model-invocation: true
+argument-hint: "<query> | --baseline | --related <exp-id>"
+allowed-tools: Read, Bash(*), Grep, Glob, WebSearch
+---
+
+Search the literature for papers, baselines, and related work.
+
+## Steps
+
+1. **Activate environment:**
+   ```bash
+   source .venv/bin/activate
+   ```
+
+2. **Parse arguments from `$ARGUMENTS`:**
+   - **Free query:** `"gradient boosting for tabular data"` — searches Semantic Scholar
+   - **Baseline:** `--baseline` — finds SOTA results for the current task, compares against your best
+   - **Related:** `--related exp-042` — finds papers using similar methods to a specific experiment
+   - `--auto-queue` — auto-queues hypotheses from literature with `source: "literature"`
+   - `--limit 10` — max number of results
+
+3. **Run literature search:**
+   ```bash
+   python scripts/literature_search.py $ARGUMENTS
+   ```
+
+4. **Report results:**
+   - **Papers:** title, authors, year, venue, citations, abstract snippet, URL
+   - **Baseline mode:** SOTA comparison with gap analysis against current best
+   - **Related mode:** methodological differences worth investigating
+   - **Hypotheses:** if `--auto-queue`, shows queued experiments from findings
+
+5. **Saved output:** results written to `experiments/literature/query-YYYY-MM-DD-HHMMSS.md`
+
+6. **If API unavailable:** reports error and suggests manual search.
+
+## Examples
+
+```
+/turing:lit "gradient boosting missing values"    # Free query
+/turing:lit --baseline                             # SOTA comparison
+/turing:lit --related exp-042                      # Related work
+/turing:lit --auto-queue "ensemble methods"        # Queue hypotheses
+```

package/skills/turing/logbook/SKILL.md

@@ -0,0 +1,51 @@
+---
+name: logbook
+description: Generate a research logbook showing the full experiment narrative — hypotheses proposed, experiments run, decisions made, and progress over time. Outputs HTML (with interactive chart) or markdown.
+disable-model-invocation: true
+argument-hint: "[--since YYYY-MM-DD] [--format html|markdown] [--output path]"
+allowed-tools: Read, Bash(python scripts/*:*, source .venv/bin/activate:*, mkdir:*), Grep, Glob
+---
+
+Generate a research logbook that captures the full narrative of the experiment campaign.
+
+## Steps
+
+1. **Generate the logbook:**
+   ```bash
+   source .venv/bin/activate && python scripts/generate_logbook.py
+   ```
+
+   **With options from `$ARGUMENTS`:**
+   - `--since 2026-03-15` — only include events after this date
+   - `--format markdown` — output as markdown instead of HTML
+   - `--output logbook.html` — write to file instead of stdout
+
+   **Common usage:**
+   ```bash
+   # HTML logbook with interactive trajectory chart
+   source .venv/bin/activate && python scripts/generate_logbook.py --output logbook.html
+
+   # Markdown for embedding in docs or READMEs
+   source .venv/bin/activate && python scripts/generate_logbook.py --format markdown --output logbook.md
+
+   # Last week's activity
+   source .venv/bin/activate && python scripts/generate_logbook.py --since 2026-03-24 --output logbook.html
+   ```
+
+2. **Present the result:**
+   - If HTML: tell the user to open the file in their browser. The logbook includes an interactive Chart.js trajectory visualization.
+   - If markdown: display inline or note the output file location.
+
+## What the Logbook Contains
+
+- **Campaign summary:** total experiments, keep rate, best metric, hypothesis count
+- **Improvement trajectory:** interactive line chart showing metric progression and best-so-far envelope
+- **Experiment log:** every experiment with ID, description, metric value, status (kept/discarded), date
+- **Hypothesis queue:** every hypothesis with source (human/agent/literature), status, priority
+
+## When to Use
+
+- To share progress with collaborators
+- Before and after meetings to show what was tried
+- To archive a completed research campaign
+- To track progress over a specific time period

package/skills/turing/merge/SKILL.md

@@ -0,0 +1,24 @@
+---
+name: merge
+description: Model merging — average weights from multiple checkpoints into a single model (soups, TIES, DARE). Free accuracy, zero latency cost.
+disable-model-invocation: true
+argument-hint: "<exp-ids...> [--method uniform|greedy|ties|dare]"
+allowed-tools: Read, Bash(*), Grep, Glob
+---
+
+Combine model weights (not predictions) into a single, better model with no latency overhead.
+
+## Steps
+
+1. **Activate environment:** `source .venv/bin/activate`
+2. **Run:** `python scripts/model_merger.py $ARGUMENTS`
+3. **Methods:** uniform soup (simple average), greedy soup (include only if improves), TIES (trim+elect+merge), DARE (drop+rescale)
+4. **Report:** compatibility check, per-model metrics, method comparison, improvement delta
+5. **Saved output:** `experiments/merges/merge-*.yaml`
+
+## Examples
+
+```
+/turing:merge exp-042 exp-053 exp-067              # All methods
+/turing:merge exp-042 exp-053 --method greedy      # Greedy soup only
+```

package/skills/turing/mode/SKILL.md

@@ -0,0 +1,43 @@
+---
+name: mode
+description: Set the research strategy mode — explore (try new things), exploit (refine what works), or replicate (verify results). Drives novelty guard policy and agent behavior.
+disable-model-invocation: true
+argument-hint: "<explore|exploit|replicate>"
+---
+
+Set the research mode for the current project. The mode determines how the novelty guard filters proposed experiments and how the agent prioritizes its work.
+
+## Modes
+
+| Mode | Novelty Guard Policy | Agent Behavior |
+|------|---------------------|----------------|
+| **explore** | Allow novel ideas, block repeats and follow-ups | Try fundamentally different approaches |
+| **exploit** | Allow follow-ups and known successes, block repeats | Refine the current best configuration |
+| **replicate** | Allow duplicate runs, block novel ideas | Re-run best experiments with different seeds |
+
+## Steps
+
+1. **Parse mode** from `$ARGUMENTS`. Must be one of: `explore`, `exploit`, `replicate`.
+
+2. **Update experiment state:**
+   ```bash
+   source .venv/bin/activate
+   python -c "
+   import yaml
+   from pathlib import Path
+   path = Path('experiment_state.yaml')
+   state = yaml.safe_load(path.read_text()) if path.exists() else {}
+   state['research_mode'] = '$ARGUMENTS'
+   path.write_text(yaml.dump(state, default_flow_style=False))
+   print(f'Research mode set to: $ARGUMENTS')
+   "
+   ```
+
+3. **Confirm** with guidance:
+   - `explore`: "The agent will prioritize novel ideas and avoid follow-ups. Best when the current approach feels exhausted."
+   - `exploit`: "The agent will refine the current best. Best when you have a promising direction."
+   - `replicate`: "The agent will re-run experiments for statistical verification. Best before declaring a winner."
+
+## Default
+
+The default mode is `exploit` (refine what works). Change to `explore` when plateauing, `replicate` before final decisions.

package/skills/turing/onboard/SKILL.md

@@ -0,0 +1,20 @@
+---
+name: onboard
+description: Project onboarding — generate a walkthrough for new collaborators. Task, history, decisions, next steps.
+disable-model-invocation: true
+argument-hint: "[--audience researcher|engineer|stakeholder] [--depth brief|full]"
+allowed-tools: Read, Bash(*), Grep, Glob
+---
+
+5-minute read that replaces a 1-hour onboarding meeting.
+
+## Steps
+1. `source .venv/bin/activate`
+2. `python scripts/generate_onboarding.py $ARGUMENTS`
+3. **Saved:** `ONBOARDING.md`
+
+## Examples
+```
+/turing:onboard
+/turing:onboard --audience engineer --depth brief
+```

package/skills/turing/paper/SKILL.md

@@ -0,0 +1,44 @@
+---
+name: paper
+description: Draft mechanical paper sections (setup, results, ablation, hyperparameters) from experiment logs. LaTeX and markdown output.
+disable-model-invocation: true
+argument-hint: "[--sections setup,results,ablation] [--format latex|markdown]"
+allowed-tools: Read, Bash(*), Grep, Glob
+---
+
+Draft paper sections directly from experiment data.
+
+## Steps
+
+1. **Activate environment:**
+   ```bash
+   source .venv/bin/activate
+   ```
+
+2. **Parse arguments from `$ARGUMENTS`:**
+   - `--sections setup,results,ablation,hyperparameters` — which sections to draft (default: all)
+   - `--format latex|markdown` — output format (default: latex)
+
+3. **Run paper drafting:**
+   ```bash
+   python scripts/draft_paper_sections.py $ARGUMENTS
+   ```
+
+4. **Report results:**
+   - **setup:** Experimental setup prose (dataset, metrics, split, seed methodology)
+   - **results:** Comparison table with all model types, best bolded, seed study stats
+   - **ablation:** Ablation table from `/turing:ablate` results
+   - **hyperparameters:** Appendix-style parameter table per model
+
+5. **Output:** Each section saved to `paper/sections/` as `.tex` or `.md`
+
+6. **Numbers are pulled directly from experiment logs** — no manual transcription needed.
+
+## Examples
+
+```
+/turing:paper                                        # All sections, LaTeX
+/turing:paper --format markdown                      # All sections, markdown
+/turing:paper --sections setup,results               # Just setup + results
+/turing:paper --sections ablation --format latex      # Just ablation table
+```

package/skills/turing/plan/SKILL.md

@@ -0,0 +1,27 @@
+---
+name: plan
+description: Research planning assistant — design a strategic experiment campaign with budget-aware ROI allocation.
+disable-model-invocation: true
+argument-hint: "[--budget 20] [--goal \"maximize F1 for production\"]"
+allowed-tools: Read, Bash(*), Grep, Glob
+---
+
+Design the next N experiments strategically, not randomly. Allocates budget by expected ROI.
+
+## Steps
+1. `source .venv/bin/activate`
+2. `python scripts/research_planner.py $ARGUMENTS`
+3. **Saved:** `experiments/plans/`
+
+## How it works
+- Analyzes experiment history to compute per-family ROI
+- Adjusts strategy priorities based on project state and goal
+- Allocates budget across: feature engineering, model search, ensemble, calibration, verification
+- Generates phased plan with specific experiment descriptions
+
+## Examples
+```
+/turing:plan --budget 20
+/turing:plan --budget 10 --goal "maximize F1 for production deployment"
+/turing:plan --budget 30 --json
+```