cs-scientist-plugin 0.1.0

package/skills/notebooklm.md

@@ -0,0 +1,95 @@
+---
+description: >-
+  Post-investigation transformation skill. Converts a cs-scientist report.md
+  into alternative formats: podcast script, FAQ, or executive briefing.
+  Never modifies the KB or the report. Always runs after Phase 10 DOCUMENT —
+  never during the research or dev loop.
+---
+
+# notebooklm
+
+## Input
+
+```
+$ARGUMENTS — path to report.md, OR topic name if report is in the active session
+```
+
+If $ARGUMENTS is a topic name, look for the report at:
+`.cs-scientist/{matching_session_id}/report.md`
+
+## Output formats
+
+Ask the user which format they want if not specified in $ARGUMENTS:
+
+```
+¿Qué formato quieres?
+A) Podcast — guión conversacional entre dos personas, ~1500 palabras
+B) FAQ — 10-15 preguntas y respuestas para alguien nuevo en el tema
+C) Briefing ejecutivo — resumen ejecutivo de 1 página, orientado a decisiones
+```
+
+### A — Podcast script
+
+Two speakers: HOST (generalist, asks questions) and EXPERT (domain knowledge).
+Structure: hook (30s) → context (2min) → 3-4 key findings (5-6min) → implications (2min) → close (30s).
+Tone: accessible, no jargon without explanation.
+
+```
+[PODCAST: {topic}]
+Based on: {report.md path}
+
+HOST: ...
+EXPERT: ...
+[continues for ~1500 words]
+
+---
+KEY SOURCES MENTIONED: {3-5 most important sources from the report}
+```
+
+### B — FAQ
+
+10-15 questions a knowledgeable non-specialist would ask.
+Derive questions from: Open Questions in KB, hypotheses, contradictions, limitations.
+Each answer cites the report section, not the original source.
+
+```
+[FAQ: {topic}]
+Based on: {report.md path}
+
+Q1: {question}
+A: {answer in 2-4 sentences, cites report section}
+
+...
+
+---
+QUESTIONS WITHOUT CLEAR ANSWERS (from KB Open Questions):
+- {question that the research did not resolve}
+```
+
+### C — Executive briefing
+
+One page. Decision-oriented. Three sections only:
+```
+[EXECUTIVE BRIEFING: {topic}]
+Based on: {report.md path} | Date: {ISO8601}
+
+SITUATION: {1 paragraph — what is known with confidence}
+IMPLICATIONS: {2-3 bullets — what this means for decisions}
+UNCERTAINTIES: {2-3 bullets — what is still unknown and why it matters}
+RECOMMENDED NEXT STEP: {1 concrete action}
+```
+
+## Rules
+
+- Every claim in the output must trace to the report — not to new information
+- Do not introduce new claims or interpretations not present in the report
+- Do not modify report.md or knowledge_base.md — read only
+- If the report has no [VERIFIED] findings yet, stop and notify:
+  `El report no tiene hallazgos verificados. Completa Phase 10 DOCUMENT antes de usar esta skill.`
+
+## NEVER
+
+- NEVER run during the research loop — only after DOCUMENT is complete
+- NEVER add new claims not in the report
+- NEVER modify the KB or the report file
+- NEVER present [HYPOTHESIS] items as confirmed findings in any output format

package/skills/paper-outline.md

@@ -0,0 +1,143 @@
+---
+description: >-
+  Converts a cs-scientist research session (knowledge_base.md + report.md) into
+  an academic paper skeleton following standard structure. Does not write the
+  paper — produces the skeleton with guidance on what goes where, derived
+  entirely from verified KB findings. Run after Phase 10 DOCUMENT is complete.
+---
+
+# paper-outline
+
+## Input
+
+```
+$ARGUMENTS — session directory path, OR topic name, OR empty to use active session
+```
+
+If $ARGUMENTS is empty or a topic name: scan `.cs-scientist/` for matching research sessions.
+
+## Pre-condition check
+
+Read `session_state.json`. If mode ≠ research or phase ≠ DOCUMENT:
+```
+Esta skill requiere una sesión de research completada (Phase 10 DOCUMENT).
+Sesión actual: modo={mode}, fase={phase}.
+Completa la investigación antes de usar paper-outline.
+```
+
+## What you read
+
+- `knowledge_base.md` — primary source. Every paper section maps to specific KB tags.
+- `report.md` (if exists) — already synthesized findings; use as guide for the Abstract and Discussion.
+
+## KB → Paper section mapping
+
+| KB tags | Paper section |
+|---------|--------------|
+| `[FACT]`, `[VERIFIED]` | Results, Background |
+| `[SYNTHESIS]` | Discussion, Conclusions |
+| `[HYPOTHESIS]` | Future Work, Limitations |
+| `[REFUTED]` | Negative Results (Appendix or Discussion) |
+| Open Questions | Future Work |
+| [DECISION] entries (if dev session mixed) | Methods |
+
+## Output format
+
+```markdown
+# Paper Outline: {topic}
+*Derivado de sesión: {session_id} | {date}*
+*Formato sugerido: {NeurIPS | ICML | Nature | arXiv preprint — inferred from topic domain}*
+
+---
+
+## Abstract (150-250 words)
+**Derivar de:** {top 3 VERIFIED findings from KB}
+Estructura obligatoria: [Problema] [Por qué importa] [Nuestro enfoque] [Resultado principal] [Implicación]
+
+Hallazgos candidatos para incluir:
+- {VERIFIED finding 1}
+- {VERIFIED finding 2}
+- {VERIFIED finding 3}
+
+---
+
+## 1. Introduction
+**Derivar de:** {problem framing from SCOPE, motivation from KB background}
+Estructura: Hook → Gap en el estado del arte → Contribución → Estructura del paper
+
+Contenido candidato de KB:
+- {relevant FACT or VERIFIED item}
+
+---
+
+## 2. Related Work
+**Derivar de:** {FACT items with citations, SYNTHESIS items connecting prior work}
+Agrupar en subsecciones por: {identified clusters from KB — e.g., "Approaches A", "Approaches B"}
+
+Fuentes KB con cita completa: {N items — list titles and years}
+Fuentes a buscar para completar: {Open Questions that require more literature}
+
+---
+
+## 3. Methods / Approach
+**Derivar de:** {DESIGN artifacts if dev, or methodology from research phases}
+Estructura: [Setup / datos] [Procedimiento] [Criterios de evaluación]
+
+---
+
+## 4. Results
+**Derivar de:** {all VERIFIED items — the confirmed findings}
+Cada subsección = un hallazgo verificado.
+
+Hallazgos confirmados para este bloque:
+{list all [VERIFIED] KB items with source citations}
+
+---
+
+## 5. Discussion
+**Derivar de:** {SYNTHESIS items, connections between findings}
+Estructura: [Lo que significan los resultados] [Por qué son sorprendentes o esperados] [Limitaciones]
+
+Candidatos de discusión de KB:
+{list [SYNTHESIS] items}
+
+---
+
+## 6. Limitations
+**Derivar de:** {HYPOTHESIS items not verified, Open Questions not resolved, [REFUTED] items that revealed scope limits}
+
+- {limitation 1 — source in KB}
+- {limitation 2 — source in KB}
+
+---
+
+## 7. Future Work
+**Derivar de:** {unresolved Open Questions, LOW-confidence HYPOTHESIS items}
+
+- {future direction 1 — linked to KB open question}
+
+---
+
+## Appendix A — Negative Results (Optional but recommended)
+**Derivar de:** {[REFUTED] items, gate failures documented in negative_results.md if available}
+Incluir si hay ≥2 hipótesis refutadas — es metodológicamente valioso.
+
+---
+
+## References
+{all cited sources from KB — grouped by section where cited}
+Formato: {inferred from target venue — APA / IEEE / Nature / ACM}
+
+---
+
+ADVERTENCIAS:
+{list of KB items marked [HYPOTHESIS] that appear in Results — these need more validation before submission}
+{list of SYNTHESIS items used as if VERIFIED — these need explicit hedging language}
+```
+
+## NEVER
+
+- NEVER include [HYPOTHESIS] items in Results without flagging them
+- NEVER omit Limitations — a paper without limitations is not credible
+- NEVER suggest a venue if you cannot match the topic domain to it with confidence
+- NEVER write the paper — only the skeleton with content guidance derived from KB

package/skills/parallel-research.md

@@ -0,0 +1,85 @@
+---
+description: >-
+  Dispatches multiple research angles as independent parallel searches.
+  Used in cs-scientist-research Phase 2 DECOMPOSE to accelerate RETRIEVE.
+  Each angle is searched independently with no cross-contamination between
+  searches. Results are consolidated in KB-compatible format.
+---
+
+# parallel-research
+
+## Input
+
+```
+$ARGUMENTS — list of research angles from DECOMPOSE, one per line
+```
+
+Format expected:
+```
+TOPIC: {main research question}
+ANGLES:
+1. {angle 1 — searchable independently}
+2. {angle 2 — searchable independently}
+3. {angle 3 — searchable independently}
+...
+```
+
+## What you do
+
+Search each angle as a fully independent research task. Do not let findings from angle N
+influence the search strategy for angle N+1. Cross-contamination defeats the purpose of
+parallel independent searches.
+
+### Per angle — search protocol
+
+For each angle:
+1. Search across ≥3 source types (academic, industry, primary data minimum)
+2. Extract and tag each claim (same tags as deep-research: `[FACT]`, `[HYPOTHESIS]`, etc.)
+3. Note contradictions within the angle
+4. Estimate coverage: `low` (<5 sources) / `medium` (5-10) / `high` (>10)
+
+### Output format
+
+```
+PARALLEL-RESEARCH: {main topic}
+ANGLES_SEARCHED: {N}
+
+---
+
+ANGLE 1: {angle name}
+COVERAGE: low | medium | high
+FACTS:
+- [FACT] {claim} — Source: {title}, {year}, {URL}
+HYPOTHESES:
+- [HYPOTHESIS] {claim} | Supporting: {source}
+CONTRADICTIONS:
+- {claim}: FOR {source} | AGAINST {source}
+OPEN_QUESTIONS:
+- {unresolved question specific to this angle}
+
+ANGLE 2: {angle name}
+...
+
+---
+
+CROSS-ANGLE PATTERNS:
+- {pattern that appears across multiple angles — mark as [SYNTHESIS] candidate}
+
+COVERAGE_GAPS:
+- {angle with low coverage that needs more research}
+```
+
+## Integration with cs-scientist-research
+
+Call this skill from Phase 2 DECOMPOSE when the angles are defined.
+Output feeds directly into Phase 3 RETRIEVE — skip manual search for covered angles.
+Low-coverage angles still require manual search in Phase 3.
+
+CROSS-ANGLE PATTERNS go into KB as `[SYNTHESIS]` candidates — they are model-generated
+connections, not verified facts. Label them explicitly.
+
+## NEVER
+
+- NEVER mix findings between angles during search — search each independently
+- NEVER mark cross-angle patterns as [FACT] — they are [SYNTHESIS] candidates until verified
+- NEVER skip low-coverage angles — flag them as gaps, do not omit them from output

package/skills/project-onboarding.md

@@ -0,0 +1,118 @@
+---
+description: >-
+  Generates a concise onboarding document for a new team member joining a
+  project. Reads AGENTS.md/CLAUDE.md, README, recent git history, and project
+  configuration files to produce a "Day 1" guide: what the project does, how to
+  run it, key architectural decisions, and where to start.
+---
+
+# project-onboarding
+
+## Input
+
+```
+$ARGUMENTS — project root path, OR empty to use current project
+```
+
+If $ARGUMENTS is empty: `git rev-parse --show-toplevel 2>/dev/null || pwd`
+
+## What you read
+
+Read these files in order. Stop reading a file after 200 lines — if it is longer, skim for
+the sections most relevant to a new developer.
+
+| File | What to extract |
+|------|----------------|
+| `AGENTS.md` or `CLAUDE.md` | Project type, stack, architecture, commands, constraints |
+| `README.md` | Purpose, setup instructions, any quickstart |
+| `package.json` / `pyproject.toml` / `Cargo.toml` / `go.mod` | Dependencies, scripts |
+| `CHANGELOG.md` | Last 3 entries — what changed recently |
+| `.gitignore` | What kind of artifacts the project generates (inferred from ignores) |
+
+Then run:
+```bash
+git log --oneline -15
+```
+to get recent commit history. Identify the 3 most active areas of the codebase.
+
+## Output format
+
+```markdown
+# Onboarding: {project name}
+*Generado el {date} desde el estado actual del repositorio*
+
+---
+
+## ¿Qué es este proyecto?
+{1-2 sentences. What problem does it solve? Who uses it?}
+
+## Stack
+
+| | |
+|-|-|
+| Lenguaje(s) | {languages} |
+| Framework(s) | {frameworks} |
+| Arquitectura | {pattern} |
+| Tipo | {web app / CLI / library / pipeline / etc.} |
+
+## Cómo arrancarlo
+
+```bash
+{exact commands from AGENTS.md or README — no guessing}
+```
+
+## Cómo testear
+
+```bash
+{exact test command}
+```
+Criterio de done: {done criterion from AGENTS.md}
+
+## Estructura del proyecto
+
+{top-level directory listing with one-line purpose per directory}
+```
+dir/        → {what lives here}
+```
+
+## Decisiones de arquitectura que necesitas conocer
+
+{from AGENTS.md "decisiones no obvias" section, or inferred from code structure}
+- {decision}: {why it was made this way}
+
+## Lo que NUNCA debes hacer
+
+{from AGENTS.md "qué NUNCA debe hacer un agente" — applies to humans too}
+- {constraint}
+
+## Actividad reciente (últimos 15 commits)
+
+{3 bullets: what areas have changed, what is actively being worked on}
+
+## Por dónde empezar
+
+{based on git log + structure: what is the main entry point? what is the core flow?}
+1. Lee: `{most important file to read first}`
+2. Ejecuta: `{command that shows the system working}`
+3. Modifica: `{smallest safe change to make to verify the dev loop works}`
+
+---
+*Este documento es una instantánea. Para el estado actual consulta AGENTS.md y git log.*
+```
+
+## Rules
+
+- Only include information derivable from the files — no fabrication
+- If a command is not in the files, write `{no encontrado — verificar con el equipo}` rather than guessing
+- The "Por dónde empezar" section must be concrete and executable, not vague
+
+## Save behavior
+
+Ask: "¿Guardo esto como `ONBOARDING.md` en el proyecto? [S/n]"
+Default yes. If yes, save to project root. If the file already exists, show a diff summary and ask before overwriting.
+
+## NEVER
+
+- NEVER invent commands — only use what appears in the files
+- NEVER include secrets, API keys, or credentials even if found in config files
+- NEVER create this file without asking first — the user decides if it should be committed

package/skills/session-status.md

@@ -0,0 +1,79 @@
+---
+description: >-
+  Shows the current state of an active cs-scientist session. Reads
+  session_state.json, goals.md, and the last 5 entries of activity_log.jsonl
+  to produce a human-readable status report. Essential after context
+  compaction or when returning to a session after a break.
+---
+
+# session-status
+
+## Input
+
+```
+$ARGUMENTS — session directory path, OR empty to scan .cs-scientist/ in current project
+```
+
+If $ARGUMENTS is empty:
+1. Run `git rev-parse --show-toplevel 2>/dev/null || pwd` to find project root
+2. List all sessions in `{project_root}/.cs-scientist/`
+3. If multiple: ask which one. If one: use it. If none: report no active session.
+
+## What you read
+
+- `session_state.json` — current phase, gate status, next action
+- `goals.md` — goal progress
+- `activity_log.jsonl` — last 5 entries (what happened recently)
+
+## Output format
+
+```
+SESSION STATUS
+──────────────────────────────────────────────
+Session:  {session_id}
+Mode:     {research | dev}
+Topic:    {topic}
+──────────────────────────────────────────────
+
+PHASE
+  Current:  {phase} ({phase_status})
+  Blocked:  {blocked_reason or "—"}
+
+GATES
+  GATE_1:      {pending | pass | fail}
+  GATE_2:      {pending | pass | fail}
+  GATE_3:      {pending | pass | fail}    ← research only
+  GATE_1_DEV:  {pending | pass | fail}    ← dev only
+  GATE_2_DEV:  {pending | pass | fail}    ← dev only
+
+GOALS
+  Active ({N}):
+    ● HIGH   {goal description}
+    ○ MEDIUM {goal description}
+  Completed ({N}): {last completed goal | "—"}
+  Blocked ({N}):   {blocked goal | "—"}
+
+RECENT ACTIVITY (last 5 actions)
+  {ts short} [{agent}] {summary}
+  {ts short} [{agent}] {summary}
+  ...
+
+NEXT ACTION
+  {next_action from session_state.json}
+
+──────────────────────────────────────────────
+Switch to cs-scientist-{research|dev} and paste:
+SESSION: {full path to session_state.json}
+TOPIC: {topic}
+NEXT_ACTION: {next_action}
+──────────────────────────────────────────────
+```
+
+The last block (the ready-to-paste dispatch input) makes resuming a session after compaction
+a single copy-paste operation.
+
+## NEVER
+
+- NEVER modify any session file — read only
+- NEVER infer state — only report what is in the files
+- NEVER omit the ready-to-paste block — it is the primary value of this skill after compaction

package/skills/writing-plans/SKILL.md

@@ -0,0 +1,152 @@
+---
+name: writing-plans
+description: Use when you have a spec or requirements for a multi-step task, before touching code
+---
+
+# Writing Plans
+
+## Overview
+
+Write comprehensive implementation plans assuming the engineer has zero context for our codebase and questionable taste. Document everything they need to know: which files to touch for each task, code, testing, docs they might need to check, how to test it. Give them the whole plan as bite-sized tasks. DRY. YAGNI. TDD. Frequent commits.
+
+Assume they are a skilled developer, but know almost nothing about our toolset or problem domain. Assume they don't know good test design very well.
+
+**Announce at start:** "I'm using the writing-plans skill to create the implementation plan."
+
+**Context:** This should be run in a dedicated worktree (created by brainstorming skill).
+
+**Save plans to:** `docs/superpowers/plans/YYYY-MM-DD-<feature-name>.md`
+- (User preferences for plan location override this default)
+
+## Scope Check
+
+If the spec covers multiple independent subsystems, it should have been broken into sub-project specs during brainstorming. If it wasn't, suggest breaking this into separate plans — one per subsystem. Each plan should produce working, testable software on its own.
+
+## File Structure
+
+Before defining tasks, map out which files will be created or modified and what each one is responsible for. This is where decomposition decisions get locked in.
+
+- Design units with clear boundaries and well-defined interfaces. Each file should have one clear responsibility.
+- You reason best about code you can hold in context at once, and your edits are more reliable when files are focused. Prefer smaller, focused files over large ones that do too much.
+- Files that change together should live together. Split by responsibility, not by technical layer.
+- In existing codebases, follow established patterns. If the codebase uses large files, don't unilaterally restructure - but if a file you're modifying has grown unwieldy, including a split in the plan is reasonable.
+
+This structure informs the task decomposition. Each task should produce self-contained changes that make sense independently.
+
+## Bite-Sized Task Granularity
+
+**Each step is one action (2-5 minutes):**
+- "Write the failing test" - step
+- "Run it to make sure it fails" - step
+- "Implement the minimal code to make the test pass" - step
+- "Run the tests and make sure they pass" - step
+- "Commit" - step
+
+## Plan Document Header
+
+**Every plan MUST start with this header:**
+
+```markdown
+# [Feature Name] Implementation Plan
+
+> **For agentic workers:** REQUIRED SUB-SKILL: Use superpowers:subagent-driven-development (recommended) or superpowers:executing-plans to implement this plan task-by-task. Steps use checkbox (`- [ ]`) syntax for tracking.
+
+**Goal:** [One sentence describing what this builds]
+
+**Architecture:** [2-3 sentences about approach]
+
+**Tech Stack:** [Key technologies/libraries]
+
+---
+```
+
+## Task Structure
+
+````markdown
+### Task N: [Component Name]
+
+**Files:**
+- Create: `exact/path/to/file.py`
+- Modify: `exact/path/to/existing.py:123-145`
+- Test: `tests/exact/path/to/test.py`
+
+- [ ] **Step 1: Write the failing test**
+
+```python
+def test_specific_behavior():
+    result = function(input)
+    assert result == expected
+```
+
+- [ ] **Step 2: Run test to verify it fails**
+
+Run: `pytest tests/path/test.py::test_name -v`
+Expected: FAIL with "function not defined"
+
+- [ ] **Step 3: Write minimal implementation**
+
+```python
+def function(input):
+    return expected
+```
+
+- [ ] **Step 4: Run test to verify it passes**
+
+Run: `pytest tests/path/test.py::test_name -v`
+Expected: PASS
+
+- [ ] **Step 5: Commit**
+
+```bash
+git add tests/path/test.py src/path/file.py
+git commit -m "feat: add specific feature"
+```
+````
+
+## No Placeholders
+
+Every step must contain the actual content an engineer needs. These are **plan failures** — never write them:
+- "TBD", "TODO", "implement later", "fill in details"
+- "Add appropriate error handling" / "add validation" / "handle edge cases"
+- "Write tests for the above" (without actual test code)
+- "Similar to Task N" (repeat the code — the engineer may be reading tasks out of order)
+- Steps that describe what to do without showing how (code blocks required for code steps)
+- References to types, functions, or methods not defined in any task
+
+## Remember
+- Exact file paths always
+- Complete code in every step — if a step changes code, show the code
+- Exact commands with expected output
+- DRY, YAGNI, TDD, frequent commits
+
+## Self-Review
+
+After writing the complete plan, look at the spec with fresh eyes and check the plan against it. This is a checklist you run yourself — not a subagent dispatch.
+
+**1. Spec coverage:** Skim each section/requirement in the spec. Can you point to a task that implements it? List any gaps.
+
+**2. Placeholder scan:** Search your plan for red flags — any of the patterns from the "No Placeholders" section above. Fix them.
+
+**3. Type consistency:** Do the types, method signatures, and property names you used in later tasks match what you defined in earlier tasks? A function called `clearLayers()` in Task 3 but `clearFullLayers()` in Task 7 is a bug.
+
+If you find issues, fix them inline. No need to re-review — just fix and move on. If you find a spec requirement with no task, add the task.
+
+## Execution Handoff
+
+After saving the plan, offer execution choice:
+
+**"Plan complete and saved to `docs/superpowers/plans/<filename>.md`. Two execution options:**
+
+**1. Subagent-Driven (recommended)** - I dispatch a fresh subagent per task, review between tasks, fast iteration
+
+**2. Inline Execution** - Execute tasks in this session using executing-plans, batch execution with checkpoints
+
+**Which approach?"**
+
+**If Subagent-Driven chosen:**
+- **REQUIRED SUB-SKILL:** Use superpowers:subagent-driven-development
+- Fresh subagent per task + two-stage review
+
+**If Inline Execution chosen:**
+- **REQUIRED SUB-SKILL:** Use superpowers:executing-plans
+- Batch execution with checkpoints for review