@mrclrchtr/supi-claude-md 0.1.0

package/skills/claude-md-improver/references/templates.md

@@ -0,0 +1,300 @@
+# CLAUDE.md Templates
+
+## Key Principles
+
+- **Concise**: Dense, human-readable content; one line per concept when possible
+- **Actionable**: Commands should be copy-paste ready
+- **Project-specific**: Document patterns unique to this project, not generic advice
+- **Current**: All info should reflect actual codebase state
+
+---
+
+## Recommended Sections
+
+Use only the sections relevant to the project. Not all sections are needed.
+
+### Commands
+
+Document the essential commands for working with the project.
+
+```markdown
+## Commands
+
+| Command | Description |
+|---------|-------------|
+| `<install command>` | Install dependencies |
+| `<dev command>` | Start development server |
+| `<build command>` | Production build |
+| `<test command>` | Run tests |
+| `<lint command>` | Lint/format code |
+```
+
+### Architecture
+
+Describe the project structure so Claude understands where things live.
+
+```markdown
+## Architecture
+
+```
+<root>/
+  <dir>/    # <purpose>
+  <dir>/    # <purpose>
+  <dir>/    # <purpose>
+```
+```
+
+### Key Files
+
+List important files that Claude should know about.
+
+```markdown
+## Key Files
+
+- `<path>` - <purpose>
+- `<path>` - <purpose>
+```
+
+### Code Style
+
+Document project-specific coding conventions.
+
+```markdown
+## Code Style
+
+- <convention>
+- <convention>
+- <preference over alternative>
+```
+
+### Environment
+
+Document required environment variables and setup.
+
+```markdown
+## Environment
+
+Required:
+- `<VAR_NAME>` - <purpose>
+- `<VAR_NAME>` - <purpose>
+
+Setup:
+- <setup step>
+```
+
+### Testing
+
+Document testing approach and commands.
+
+```markdown
+## Testing
+
+- `<test command>` - <what it tests>
+- <testing convention or pattern>
+```
+
+### Gotchas
+
+Document non-obvious patterns, quirks, and warnings.
+
+```markdown
+## Gotchas
+
+- <non-obvious thing that causes issues>
+- <ordering dependency or prerequisite>
+- <common mistake to avoid>
+```
+
+### Workflow
+
+Document development workflow patterns.
+
+```markdown
+## Workflow
+
+- <when to do X>
+- <preferred approach for Y>
+```
+
+---
+
+## Template: Project Root (Minimal)
+
+```markdown
+# <Project Name>
+
+<One-line description>
+
+## Commands
+
+| Command | Description |
+|---------|-------------|
+| `<command>` | <description> |
+
+## Architecture
+
+```
+<structure>
+```
+
+## Gotchas
+
+- <gotcha>
+```
+
+---
+
+## Template: Project Root (Comprehensive)
+
+```markdown
+# <Project Name>
+
+<One-line description>
+
+## Commands
+
+| Command | Description |
+|---------|-------------|
+| `<command>` | <description> |
+
+## Architecture
+
+```
+<structure with descriptions>
+```
+
+## Key Files
+
+- `<path>` - <purpose>
+
+## Code Style
+
+- <convention>
+
+## Environment
+
+- `<VAR>` - <purpose>
+
+## Testing
+
+- `<command>` - <scope>
+
+## Gotchas
+
+- <gotcha>
+```
+
+---
+
+## Template: Package/Module
+
+For packages within a monorepo or distinct modules.
+
+```markdown
+# <Package Name>
+
+<Purpose of this package>
+
+## Usage
+
+```
+<import/usage example>
+```
+
+## Key Exports
+
+- `<export>` - <purpose>
+
+## Dependencies
+
+- `<dependency>` - <why needed>
+
+## Notes
+
+- <important note>
+```
+
+---
+
+## Template: Monorepo Root
+
+```markdown
+# <Monorepo Name>
+
+<Description>
+
+## Packages
+
+| Package | Description | Path |
+|---------|-------------|------|
+| `<name>` | <purpose> | `<path>` |
+
+## Commands
+
+| Command | Description |
+|---------|-------------|
+| `<command>` | <description> |
+
+## Cross-Package Patterns
+
+- <shared pattern>
+- <generation/sync pattern>
+```
+
+> ⚠️ **When SuPi is active:** The `## Packages` table in this template is auto-generated by `code_intel brief` on the first turn. Consider omitting it or replacing it with `## Start Here` and `## Cross-Package Patterns` sections that capture conventions not visible in manifests.
+
+---
+
+## Template: SuPi-Optimized Project Root
+
+For projects using SuPi extensions (`code-intelligence`, `claude-md`). Omits sections that extensions auto-deliver and favors compressed human-only guidance after a baseline overlap review.
+
+```markdown
+# <Project Name>
+
+<One-line description — what makes this project unique>
+
+## Commands
+
+| Command | Description |
+|---------|-------------|
+| `<command>` | <description> |
+
+## Start Here
+
+- <where to begin for the most common task>
+- <entry point or file to read first>
+
+## Cross-Package Patterns
+
+- <pattern that isn't obvious from code>
+- <convention that required a decision>
+
+## Gotchas
+
+- <non-obvious thing that causes issues>
+- <ordering dependency or prerequisite>
+
+## Workflow
+
+- <when to do X>
+- <preferred approach for Y>
+```
+
+**What's intentionally absent:**
+- `## Modules` / `## Packages` — `code_intel brief` generates this
+- Large root `## Project structure` / `## Architecture` directory trees that just restate the workspace layout
+- `## Dependencies` table — derivable from `package.json`
+
+If you keep any structure section in a SuPi-enabled root doc, make it short and opinionated: explain where to start, why boundaries exist, or which exception matters. After the baseline review, split the redundant tree/inventory portion from the unique guidance portion instead of keeping both together. Do not paste the tree unless the tree itself carries human-only meaning.
+
+---
+
+## Update Principles
+
+When updating any CLAUDE.md:
+
+1. **Be specific**: Use actual file paths, real commands from this project
+2. **Be current**: Verify info against the actual codebase
+3. **Be brief**: One line per concept when possible
+4. **Be useful**: Would this help a new Claude session understand the project?

package/skills/claude-md-improver/references/update-guidelines.md

@@ -0,0 +1,218 @@
+# CLAUDE.md Update Guidelines
+
+## Core Principle
+
+Only add information that will genuinely help future Claude sessions. The context window is precious - every line must earn its place.
+
+## What TO Add
+
+### 1. Commands/Workflows Discovered
+
+```markdown
+## Build
+
+`npm run build:prod` - Full production build with optimization
+`npm run build:dev` - Fast dev build (no minification)
+```
+
+Why: Saves future sessions from discovering these again.
+
+### 2. Gotchas and Non-Obvious Patterns
+
+```markdown
+## Gotchas
+
+- Tests must run sequentially (`--runInBand`) due to shared DB state
+- `yarn.lock` is authoritative; delete `node_modules` if deps mismatch
+```
+
+Why: Prevents repeating debugging sessions.
+
+### 3. Package Relationships
+
+```markdown
+## Dependencies
+
+The `auth` module depends on `crypto` being initialized first.
+Import order matters in `src/bootstrap.ts`.
+```
+
+Why: Architecture knowledge that isn't obvious from code.
+
+### 4. Testing Approaches That Worked
+
+```markdown
+## Testing
+
+For API endpoints: Use `supertest` with the test helper in `tests/setup.ts`
+Mocking: Factory functions in `tests/factories/` (not inline mocks)
+```
+
+Why: Establishes patterns that work.
+
+### 5. Configuration Quirks
+
+```markdown
+## Config
+
+- `NEXT_PUBLIC_*` vars must be set at build time, not runtime
+- Redis connection requires `?family=0` suffix for IPv6
+```
+
+Why: Environment-specific knowledge.
+
+## What SuPi Extensions Already Deliver
+
+When auditing or updating CLAUDE.md in a project using SuPi extensions, these sections are redundant because they're auto-injected on every session:
+
+| Extension | What It Injects | When |
+|-----------|----------------|------|
+| `supi-code-intelligence` | Workspace module graph (names, descriptions, paths, dependency edges) | First `before_agent_start` per session |
+| `supi-claude-md` | Subdirectory `CLAUDE.md` files wrapped in `<extension-context>` | On `read`/`edit`/`lsp`/`tree_sitter` to subdirectories |
+| `supi-core` | `findProjectRoot`, `walkProject`, XML `<extension-context>` tagging | Available to all extensions |
+
+**Implication:** A root CLAUDE.md doesn't need to document what `code_intel brief` would say. Focus instead on:
+- Commands and workflows
+- Cross-package conventions and patterns
+- Gotchas that aren't in code
+- Human-curated guidance ("start here for X")
+
+When SuPi is active, do a quick baseline review first: compare the CLAUDE.md against `code_intel brief` and other known injected context, then separate each section into overlap vs unique value. If a root `## Project structure` / `## Architecture` section mostly restates the workspace tree, treat that portion as redundant and keep only the orientation, boundary rules, or exceptions that the generated overview cannot supply.
+
+## What NOT to Add
+
+### 1. Obvious Code Info
+
+Bad:
+```markdown
+The `UserService` class handles user operations.
+```
+
+The class name already tells us this.
+
+### 2. Generic Best Practices
+
+Bad:
+```markdown
+Always write tests for new features.
+Use meaningful variable names.
+```
+
+This is universal advice, not project-specific.
+
+### 3. One-Off Fixes
+
+Bad:
+```markdown
+We fixed a bug in commit abc123 where the login button didn't work.
+```
+
+Won't recur; clutters the file.
+
+### 4. Verbose Explanations
+
+Bad:
+```markdown
+The authentication system uses JWT tokens. JWT (JSON Web Tokens) are
+an open standard (RFC 7519) that defines a compact and self-contained
+way for securely transmitting information between parties as a JSON
+object. In our implementation, we use the HS256 algorithm which...
+```
+
+Good:
+```markdown
+Auth: JWT with HS256, tokens in `Authorization: Bearer <token>` header.
+```
+
+## Diff Format for Updates
+
+For each suggested change:
+
+### 1. Identify the File
+
+```
+File: ./CLAUDE.md
+Section: Commands (new section after ## Architecture)
+```
+
+### 2. Show the Change
+
+```diff
+ ## Architecture
+ ...
+
++## Commands
++
++| Command | Purpose |
++|---------|---------|
++| `npm run dev` | Dev server with HMR |
++| `npm run build` | Production build |
++| `npm test` | Run test suite |
+```
+
+### 3. Explain Why
+
+> **Why this helps:** The build commands weren't documented, causing
+> confusion about how to run the project. This saves future sessions
+> from needing to inspect `package.json`.
+
+## What NOT to Add (SuPi Projects)
+
+In addition to the existing guidelines, avoid these when SuPi is active:
+
+**Redundant: Package/module inventory**
+```markdown
+Bad:
+## Packages
+| Package | Description | Path |
+|---------|-------------|------|
+| `api` | REST API | `packages/api/` |
+
+Better: Skip entirely, or if relationships are non-obvious:
+## Cross-Package Patterns
+The `api` package must be initialized before `worker` due to shared DB migrations.
+```
+
+**Redundant: High-level dependency graph**
+```markdown
+Bad:
+## Dependencies
+- `api` depends on `db`, `auth`
+- `web` depends on `api`
+
+Better: Skip — `code_intel brief` shows this live.
+```
+
+**Partially redundant: Root project structure section with both overlap and unique value**
+```markdown
+Overlap portion:
+## Project structure
+- `apps/web` — frontend
+- `apps/api` — backend
+- `packages/db` — shared database code
+- `packages/ui` — shared components
+
+Keep portion:
+- `packages/db` owns schema changes; app packages consume generated clients only
+- API request flow starts at `apps/api/src/routes/` and drops into `packages/db/`
+
+Better rewrite:
+## Start Here
+- Web changes usually start in `apps/web/src/app/`
+- API request flow starts at `apps/api/src/routes/` and drops into `packages/db/`
+
+## Cross-Package Patterns
+- `packages/db` owns schema changes; app packages consume generated clients only
+```
+
+## Validation Checklist
+
+Before finalizing an update, verify:
+
+- [ ] Each addition is project-specific
+- [ ] No generic advice or obvious info
+- [ ] Commands are tested and work
+- [ ] File paths are accurate
+- [ ] Would a new Claude session find this helpful?
+- [ ] Is this the most concise way to express the info?
+- [ ] No overlap with SuPi auto-delivered content (when SuPi is active)

package/skills/claude-md-revision/SKILL.md

@@ -0,0 +1,126 @@
+---
+name: claude-md-revision
+description: "Update CLAUDE.md, .claude.local.md, or AGENTS.md files with project-specific context for future agent sessions. Trigger ONLY when the user explicitly mentions CLAUDE.md, .claude.local.md, AGENTS.md, project memory, or agent context files. Also trigger on direct phrases like 'add this to CLAUDE.md', 'remember this for next time', 'update project memory', or 'document this pattern'. Do NOT trigger for general documentation requests, bug reports, session summaries, or user-facing docs — those are unrelated to agent instruction files."
+license: MIT
+---
+
+# Revise CLAUDE.md with Session Learnings
+
+Review the current session for actionable learnings and update the project's CLAUDE.md or .claude.local.md files. The goal is concise, durable context that helps future sessions work more effectively in this codebase.
+
+## Workflow
+
+### Step 1: Reflect
+
+Identify what context would have helped this session go smoother. Scan the conversation for learnings worth preserving.
+
+**What TO capture:**
+- **Commands/workflows** discovered or used repeatedly
+- **Gotchas and non-obvious patterns** (deprecated APIs, common pitfalls, tricky behaviors)
+- **Package relationships** or ordering dependencies not obvious from code
+- **Testing approaches** that worked (or didn't)
+- **Configuration quirks** (env vars, build tools, path oddities)
+
+**What NOT to capture:**
+- Obvious info already clear from code or naming
+- Generic best practices ("write tests", "use meaningful names")
+- One-off fixes unlikely to recur ("fixed typo on line 42")
+- Verbose explanations — one line per concept; link to docs for detail
+- Content that SuPi extensions already auto-deliver (package tables, module graphs, dependency lists from manifests — see update guidelines)
+
+Filter ruthlessly. CLAUDE.md is part of the prompt — brevity directly improves future sessions.
+
+> See also: [detailed update guidelines](references/update-guidelines.md) and [quality criteria](references/quality-criteria.md)
+
+### Step 2: Find Context Files
+
+Find existing CLAUDE.md and .claude.local.md files in the project:
+
+```bash
+find . -name "CLAUDE.md" -o -name ".claude.local.md" 2>/dev/null | head -20
+```
+
+For each file found, decide if the new content belongs there:
+
+| File | Purpose |
+|------|---------|
+| `CLAUDE.md` | Team-shared context. Check into git. Use for patterns, commands, and gotchas that any teammate would benefit from. |
+| `.claude.local.md` | Personal/local only. Gitignored. Use for personal preferences, local environment quirks, or experimental notes. |
+
+If no CLAUDE.md exists at the project root and the learnings are team-relevant, create one.
+
+### Step 3: Draft Additions
+
+**Keep it concise** — one line per concept. Use this format:
+
+```
+<command or pattern> — <brief description>
+```
+
+Examples:
+
+```
+`pnpm vitest run packages/<pkg>/` — run tests for a single workspace package
+`os.homedir()` cannot be mocked in ESM — pass `homeDir` parameter for testability
+Root context files are never re-injected by extensions; use `/reload` to refresh
+```
+
+Avoid:
+- Verbose explanations (link to docs if detail is needed)
+- Obvious information ("JavaScript uses `const` for constants")
+- One-off fixes unlikely to recur ("Fixed typo in variable name on line 42")
+- Duplicating existing content (read the file first)
+
+### Step 4: Propose Changes
+
+For each addition, present it in this format:
+
+```markdown
+### Update: ./path/to/CLAUDE.md
+
+**Why:** [one-line reason this belongs here]
+
+```diff
++ [the addition — keep it brief]
+```
+```
+
+Group related additions under a single heading. If multiple files need updates, show each separately.
+
+**Diff structure:**
+1. **Identify the file and section** — existing heading, or note if new
+2. **Show the change** — concise diff or quoted block
+3. **Explain why** — one sentence on how this helps future sessions
+
+### Step 5: Apply with Approval
+
+Ask the user if they want to apply the changes. Only edit files they approve.
+
+If the user approves:
+- Append new content in the appropriate section, or create the section if it doesn't exist
+- Maintain alphabetical or logical ordering within sections
+- Run `git diff` (or equivalent) after editing to confirm the change looks correct
+
+If the user rejects or modifies a proposal, adjust and re-present.
+
+**Validation checklist** — before finalizing, verify:
+- [ ] Each addition is project-specific, not generic advice
+- [ ] No duplication with existing content (read the file first)
+- [ ] Commands are tested and work
+- [ ] File paths are accurate
+- [ ] Team-shared vs. personal-local scope is respected
+- [ ] This is the most concise way to express the info
+
+## Guidelines
+
+- **Brevity is the priority** — CLAUDE.md content is injected into every prompt. Long files hurt performance.
+- **Actionability** — each line should save future sessions time or prevent a mistake.
+- **No duplication** — read the existing file before adding; merge with existing entries when possible.
+- **Respect scope** — team-shared vs. personal-local. Don't put machine-specific paths in CLAUDE.md.
+- **Format consistency** — match the existing format in the target file. If the file uses backticks for commands, use backticks. If it uses bullet points, use bullet points.
+
+## References
+
+- [Update guidelines](references/update-guidelines.md) — detailed rubric for what to add and what to skip
+- [Quality criteria](references/quality-criteria.md) — scoring rubric for assessing CLAUDE.md quality
+- [Templates](references/templates.md) — section templates for creating new CLAUDE.md files

package/skills/claude-md-revision/evals/evals.json

@@ -0,0 +1,43 @@
+{
+  "skill_name": "claude-md-revision",
+  "evals": [
+    {
+      "id": 1,
+      "prompt": "I just spent way too long figuring out that `pnpm exec biome check --write --unsafe` is the only way to auto-fix unused imports in this repo. The regular `--write` doesn't catch them. Can you add this to the project memory so I don't waste time on it again?",
+      "expected_output": "Updates CLAUDE.md with a concise entry about biome's --unsafe flag for fixing unused imports, presented as a diff for approval",
+      "files": [],
+      "expectations": [
+        "The assistant finds the project's CLAUDE.md file",
+        "The proposed addition is concise (one line)",
+        "The addition uses the format `command` — description",
+        "The assistant asks for approval before editing",
+        "The addition is specific and actionable"
+      ]
+    },
+    {
+      "id": 2,
+      "prompt": "This session taught me that `os.homedir()` can't be mocked in ESM, so we have to pass `homeDir` as a parameter to config functions for testability. Also, `ctx.ui.theme` doesn't expose an 'info' color — you have to use 'accent' or 'dim' instead. Update the project docs with these gotchas.",
+      "expected_output": "Updates CLAUDE.md with two concise entries covering the os.homedir() mocking issue and the missing 'info' color in ctx.ui.theme",
+      "files": [],
+      "expectations": [
+        "Both gotchas are captured",
+        "Each entry is one line",
+        "Entries are formatted consistently",
+        "The assistant asks for approval before editing",
+        "No verbose explanations are added"
+      ]
+    },
+    {
+      "id": 3,
+      "prompt": "Remember for next time: the pre-push hook in this repo runs `pnpm verify`, which includes both linting and tests. Don't try to run them separately.",
+      "expected_output": "Updates CLAUDE.md with a brief note about the pre-push hook running pnpm verify",
+      "files": [],
+      "expectations": [
+        "The addition mentions the pre-push hook",
+        "The addition mentions `pnpm verify`",
+        "The entry is concise",
+        "The assistant asks for approval before editing"
+      ]
+    }
+  ]
+}