Kyle's Claude Code Plugins

Personal plugin marketplace for Claude Code workflow extensions with beads/deciduous tracking integration, lightly and lovingly wrapping ed3d's coding plugins. In desperate search of a name that doesn't suck.

Installation

1. Register the marketplace

Claude Code:

/plugin marketplace add kylestratis/kyle-claude-plugins

OMP:

omp plugin marketplace add kylestratis/kyle-claude-plugins

Use a local path instead of the repository slug to install from a working tree: add /path/to/kyle-claude-plugins. Installs then resolve from that checkout, including uncommitted changes.

2. Install plugins

The tracking hooks need their scripts executable before installation:

chmod +x plugins/tracking-hooks/hooks/*.sh plugins/tracking-hooks/hooks/*.py

Claude Code:

# Workflow commands (required)
/plugin install workflow-commands@kyle-claude-plugins

# Tracking hooks (optional - session start reminders, git hooks)
/plugin install tracking-hooks@kyle-claude-plugins

# Documentation review (optional - prose review with controlled fixes)
/plugin install documentation-review@kyle-claude-plugins

OMP:

omp plugin install --scope user workflow-commands@kyle-claude-plugins
omp plugin install --scope user tracking-hooks@kyle-claude-plugins
omp plugin install --scope user documentation-review@kyle-claude-plugins

Use --scope project to install for one repository instead of your user account.

3. Load the plugins

Run /reload-plugins in either runtime, or start a new session.

Updating

OMP upgrades in place. Claude Code pins each install to a commit, so reinstall to pick up new commits:

# OMP
omp plugin upgrade documentation-review@kyle-claude-plugins

# Claude Code
/plugin marketplace update kyle-claude-plugins
/plugin uninstall documentation-review@kyle-claude-plugins
/plugin install documentation-review@kyle-claude-plugins

Prerequisites

Required Tools

1. beads (Task Tracking)

Local task/issue tracking stored in .beads/ directory.

# Install
curl -fsSL https://raw.githubusercontent.com/steveyegge/beads/main/scripts/install.sh | bash

# Verify
bd --version

Key commands:

bd init                          # Initialize in project
bd create "Task" -t task -p 3    # Create task (priority 0-4)
bd list                          # List all tasks
bd ready                         # Show unblocked tasks
bd update <id> --status done     # Update status
bd close <id> --reason "Done"    # Close task

2. deciduous (Decision Journaling)

Decision graph stored in .deciduous/ directory.

# Install
npm install -g deciduous

# Verify
deciduous --version

Key commands:

deciduous init                             # Initialize in project
deciduous add goal "What I'm doing" -c 80  # Confidence 0-100
deciduous add decision "Choice and why" -c 80
deciduous add action "What I did"
deciduous add outcome "What happened"
deciduous add action "Committed X" --commit HEAD  # Link to git
deciduous query                            # View graph

3. ed3d-plugins (Core Workflow)

Design and implementation workflow framework.

# Register marketplace
/plugin marketplace add https://github.com/ed3dai/ed3d-plugins.git

# Install plugins
/plugin install ed3d-plan-and-execute@ed3d-plugins
/plugin install ed3d-research-agents@ed3d-plugins
/plugin install ed3d-basic-agents@ed3d-plugins

Optional Tools

Linear Integration

If using the --linear flag with /workflow-commands:intake, or passing Linear references to /workflow-commands:orchestrate:

Must have Linear MCP connector enabled in Claude
Requires access to at least one Linear team

GitHub Codex Review

The /workflow-commands:fix-pr-review command requires:

GitHub CLI authenticated through gh auth login
An open GitHub pull request
Codex code review enabled for the repository

The workflow uses GitHub CLI authentication and never reads or prints token values.

Pre-commit

For automatic verification hooks:

pip install pre-commit
pre-commit install  # In each project

Commands Reference

`/workflow-commands:project-init`

Initialize beads and deciduous tracking in a project.

/workflow-commands:project-init [--beads-only | --deciduous-only] [--with-guidance]

Flag	Description
`--beads-only`	Only initialize beads, skip deciduous
`--deciduous-only`	Only initialize deciduous, skip beads
`--with-guidance`	Create `.ed3d/` directory with customization files

Creates:

.beads/ - Task database
.deciduous/ - Decision graph
.ed3d/ - Guidance files (if --with-guidance)

`/workflow-commands:intake`

Import features from a roadmap file or inline list.

# From markdown file
/workflow-commands:intake @docs/roadmap.md

# With Linear sync
/workflow-commands:intake @docs/roadmap.md --linear

# Inline features
/workflow-commands:intake "OAuth integration" "Rate limiting" "Audit logging"

# Inline with Linear and team
/workflow-commands:intake "Feature 1" "Feature 2" --linear --team Engineering

Argument	Description
`@file.md`	Path to markdown roadmap file
`"feature"`	Inline feature names (quoted)
`--linear`	Also create Linear issues
`--team <n>`	Linear team name (required if multiple teams)

Roadmap format:

## Q1 Features

### OAuth Integration
Support Google and GitHub providers.
Priority: High

### Rate Limiting  
Token bucket algorithm.
Priority: Normal

Creates:

Beads epic per feature (with priority)
Linear issue per feature (if --linear)
Bidirectional links between beads and Linear

`/workflow-commands:explore`

Research a topic with automatic tracking.

/workflow-commands:explore <topic> [--codebase-only | --web-only] [--no-tracking]

Flag	Description
`--codebase-only`	Only search local codebase
`--web-only`	Only search web
`--no-tracking`	Skip beads task creation (deciduous still logs)

Examples:

/workflow-commands:explore OAuth2 PKCE flow
/workflow-commands:explore authentication patterns --codebase-only
/workflow-commands:explore React 19 changes --web-only

Tracking:

Creates deciduous goal at start
Creates beads task (unless --no-tracking)
Logs outcome on completion
Closes beads task

`/workflow-commands:design`

Start a design plan with decision logging. Wraps ed3d-plan-and-execute:start-design-plan.

/workflow-commands:design [feature description]

Tracking:

Creates beads epic: "Design: <feature>"
Logs deciduous goal
Logs every design decision during brainstorming
Updates beads on completion

Decision logging examples:

deciduous add decision "Chose PostgreSQL over MongoDB because: ACID compliance required, team expertise, JOIN performance" -c 85
deciduous add decision "Using event sourcing pattern because: audit trail required, temporal queries needed" -c 80

`/workflow-commands:plan`

Create an implementation plan with task creation. Wraps ed3d-plan-and-execute:start-implementation-plan.

/workflow-commands:plan @path/to/design.md <workdir>

Argument	Description
`@design.md`	Path to design document
`workdir`	Working directory (usually `.`)

Example:

/clear
/workflow-commands:plan @docs/design-plans/2026-02-10-oauth.md .

Tracking:

Creates beads task per implementation phase
Sets up task dependencies (sequential)
Logs planning decisions to deciduous

`/workflow-commands:execute`

Execute an implementation plan with status updates. Wraps ed3d-plan-and-execute:execute-implementation-plan.

/workflow-commands:execute <plan-dir> <workdir>

Argument	Description
`plan-dir`	Path to implementation plan directory
`workdir`	Working directory (usually `.`)

Example:

/clear
/workflow-commands:execute docs/implementation-plans/2026-02-10-oauth/ .

Tracking:

Updates beads task status per phase (in_progress → done)
Logs implementation decisions as they happen
Uses LEARNED pattern for discoveries

Decision logging examples:

deciduous add decision "Implemented retry with exponential backoff because upstream API is flaky" -c 80
deciduous add decision "Resolved circular dependency by extracting shared types to common module" -c 85

`/workflow-commands:verify`

Final verification with intelligent tooling detection.

/workflow-commands:verify [--skip-tests] [--skip-lint] [--skip-precommit] [--skip-review] [--task <id>]

Flag	Description
`--skip-tests`	Skip test execution
`--skip-lint`	Skip linting
`--skip-precommit`	Skip pre-commit hooks
`--skip-review`	Skip code review
`--task <id>`	Beads task ID to close on success

Automatic detection:

Language	Tests	Lint	Format
Python	pytest, tox, nox, unittest	ruff, flake8, pylint	black, ruff format
JS/TS	npm test (jest, vitest, mocha)	eslint	prettier
Rust	cargo test	cargo clippy	cargo fmt
Go	go test	golangci-lint, go vet	gofmt

Also detects:

Pre-commit hooks (.pre-commit-config.yaml)
Type checkers (mypy, tsc)
Makefile targets
Package.json scripts

Behavior:

Detects all project tooling
Runs tests, linters, formatters, pre-commit
Attempts auto-fix for formatting issues
Dispatches code review
Logs deciduous outcome
Closes beads task (if --task provided)

`/workflow-commands:fix-pr-review`

Process Codex Review findings on an open GitHub pull request.

# Current branch's PR
/workflow-commands:fix-pr-review

# Explicit PR
/workflow-commands:fix-pr-review 123
/workflow-commands:fix-pr-review https://github.com/owner/repo/pull/123

The workflow waits for Codex Review, evaluates each finding against the current code, fixes legitimate issues, and replies with technical evidence when a finding is invalid. After fixes are pushed, it requests another Codex review and repeats until the current PR head has no actionable findings.

/verify remains the pre-PR local verification workflow. /fix-pr-review operates on asynchronous Codex findings posted through the GitHub PR interface.

`/workflow-commands:pollinate`

Port a feature from another codebase into the current project.

/workflow-commands:pollinate <source-path-or-url> <feature> [--rigor critical]

Argument	Description
`<source-path-or-url>`	Local path or URL of the source codebase
`<feature>`	Feature to port
`--rigor critical`	Raise verification rigor for high-risk ports

Analyzes the source feature, maps it to this project's conventions, and generates a design document. It does not write code directly — it feeds the standard /plan → /execute → /verify pipeline.

Examples:

/workflow-commands:pollinate ../other-repo "retry middleware"
/workflow-commands:pollinate https://github.com/owner/repo "token refresh" --rigor critical

`/workflow-commands:pollinate-verify`

Three-layer verification for ported code.

/workflow-commands:pollinate-verify [--task <beads-id>] [--source <path>]

Argument	Description
`--task`	Beads task to close on success
`--source`	Source codebase to compare against

Runs after /execute completes a pollination port: differential tests against the source, adversarial hardening, then standard project verification.

`/workflow-commands:task`

Small standalone work that doesn't need design/planning phases.

/workflow-commands:task "<task description>" [--priority <0-4>]

Argument	Description
`"<description>"`	Task description (quoted)
`--priority`	0=Critical, 1=High, 2=Medium, 3=Normal (default), 4=Backlog

Use when: Work is clear, takes <1 hour, doesn't require architectural decisions.

Examples:

/workflow-commands:task "Add logging to auth middleware"
/workflow-commands:task "Update error message for invalid email" --priority 2

`/workflow-commands:bug`

Fix a bug with investigation and root cause tracking.

/workflow-commands:bug "<bug description>" [--priority <0-4>] [--blocks <task-id>]

Argument	Description
`"<description>"`	Bug description (quoted)
`--priority`	0=Critical, 1=High, 2=Medium (default), 3=Normal, 4=Backlog
`--blocks`	Link to task this bug is blocking

Workflow:

Creates beads bug
Investigate → find root cause
Fix → verify
Close bug

Examples:

/workflow-commands:bug "Login button doesn't respond on mobile"
/workflow-commands:bug "Users can't checkout" --priority 1
/workflow-commands:bug "Auth crashes on empty token" --blocks beads-a1b2

`/workflow-commands:continue`

Resume work on an existing beads task.

/workflow-commands:continue [<task-id>]

Argument	Description
`<task-id>`	Beads task ID (e.g., `beads-a1b2` or just `a1b2`)

If no task ID: Shows available tasks (bd ready, bd list --status=in_progress).

Examples:

/workflow-commands:continue beads-a1b2  # Resume specific task
/workflow-commands:continue              # Show available tasks

`/workflow-commands:orchestrate`

Coordinate independent open leaf Beads tickets in isolated worktrees. Linear issues can be passed directly.

/workflow-commands:orchestrate [<beads-id | linear-ref> ...] [--max-parallel <count>]

Argument	Description
`<beads-id>`	Restrict the candidate set to specific open leaf Beads tickets
`<linear-ref>`	A Linear issue identifier such as `ENG-123`, or a `linear.app` issue URL
`--max-parallel <count>`	Optional exact two-token positive worker limit; use it at most once; defaults to 4 and is capped at the runtime limit

Linear references require the Linear MCP connector. Each one resolves to exactly one Beads ticket by linear.identifier metadata, or creates that ticket from the Linear issue when no mapping exists, before any claim.

The two trackers have fixed roles. Beads is the execution tracker that agents read and write; it holds claims, dependency order, ownership, and completion evidence. Linear is the human progress view: it is mirrored to the started state at claim, and to the completed state with commit or PR evidence only after reconciliation is final. Provisional closures are never mirrored, a rejected reconciliation posts a supersession comment, and a mirror failure never changes Beads state, blocks the run, or counts as evidence.

Without ticket IDs, the workflow starts from bd ready. Each invocation creates a new run and rejects in-progress, blocked, or stale-owned tickets; useful state stays on its exact branch and worktree for explicit manual recovery, while an interrupted empty claim requires administrative release. The task runtime owns its configured automatic provider fallback chain. The workflow creates its integration branch and all worker branches from one target revision, keeps all tracker state in the integration worktree, integrates worker commits serially, and verifies source content while excluding the exact tracked export allowlist. It offers a local merge, a pull request, or keep-as-is; accepted worker work cannot be discarded. Ticket comments, closures, and outcomes become final only after the exact tracker-only reconciliation commit passes its path check and a fresh fetch proves that the target branch contains it; pushing a reconciliation branch or opening its pull request is not finality.

`/documentation-review:review`

Review repository documentation, prompts, docstrings, and comments against evidence-backed writing rules.

/documentation-review:review [--scope pr|repository] [--path <path> ...] \
                             [--profile <path>=<profile> ...] \
                             [--apply-approved <finding-id> ... | --autofix]

Argument	Description
`--scope`	`pr` (default) reviews prose changed from the verified PR base; `repository` reviews all eligible prose
`--path`	Repeated, repository-relative; narrows either scope. Cannot re-include excluded content
`--profile`	Repeated `<path>=<profile>`; overrides content classification
`--apply-approved`	Repeated `DR-###` IDs from the immediately preceding review in this conversation
`--autofix`	Applies all and only `safe` findings. Mutually exclusive with `--apply-approved`

Profiles: general, documentation, procedure, prompt, docstring, comment, historical-record.

Fix safety classes:

Class	Behavior
`safe`	Applied by `--autofix` or when approved
`review-required`	Applied only when explicitly approved by ID
`report-only`	Never applied automatically, even when selected

Default invocation modifies no files: it reports findings and creates a review snapshot that --apply-approved consumes. Every edit re-reads its target and requires byte-for-byte equality with the reviewed evidence, so a concurrently changed target stops that edit alone and reports a recovery action. Protected content — commands, identifiers, literals, URLs, quoted text, and necessary provenance — is never rewritten.

Examples:

# Review prose changed in the current PR
/documentation-review:review

# Review the whole repository, narrowed to two paths
/documentation-review:review --scope repository --path README.md --path docs/

# Treat a file as a procedure regardless of content classification
/documentation-review:review --profile docs/deploy.md=procedure

# Apply two approved findings from the preceding review
/documentation-review:review --apply-approved DR-001 --apply-approved DR-004

# Apply only findings classified safe
/documentation-review:review --scope repository --autofix

Entry Points

Start From	When to Use	Creates
`/workflow-commands:design`	New feature needing architecture	Epic
`/workflow-commands:plan`	Have design doc, need tasks	Phase tasks
`/workflow-commands:task`	Small standalone work (<1hr)	Single task
`/workflow-commands:bug`	Fix a bug	Bug
`/workflow-commands:continue`	Resume existing work	Nothing (uses existing)
`/workflow-commands:orchestrate`	Run independent ready tickets concurrently	Nothing (uses existing)
`/workflow-commands:explore`	Research before design	Research task
`/workflow-commands:intake`	Import roadmap	Epics
`/workflow-commands:pollinate`	Port a feature from another codebase	Design document
`/documentation-review:review`	Review prose before commit or PR	Nothing (report + review snapshot)

Workflow

Full Feature Workflow

/workflow-commands:project-init          # Once per project
    │
    ▼
/workflow-commands:intake @roadmap.md --linear   # Import features
    │
    │   ┌───────────────────────────────────────┐
    │   │  For each feature:                    │
    ▼   ▼                                       │
/workflow-commands:explore <topic>              │  ← deciduous: goal/outcome
    │                                           │    beads: research task
    ▼                                           │
/workflow-commands:design                       │  ← deciduous: ALL design decisions
    │                                           │    beads: design epic
    ▼                                           │
/clear                                          │
/workflow-commands:plan @design.md .            │  ← deciduous: planning decisions
    │                                           │    beads: phase tasks
    ▼                                           │
/clear                                          │
/workflow-commands:execute <plan-dir> .         │  ← deciduous: impl decisions
    │                                           │    beads: status updates
    ▼                                           │
/workflow-commands:verify --task <epic-id>      │  ← deciduous: outcome
    │                                           │    beads: close epic
    ▼                                           │
git commit / PR ────────────────────────────────┘

Parallel Ticket Workflow

For existing ready tickets with disjoint ownership:

/workflow-commands:orchestrate beads-a1b2 beads-c3d4 --max-parallel 2
/workflow-commands:orchestrate ENG-412 ENG-419   # Linear issues, resolved to beads tickets first
# isolated workers → serial integration → combined verification → publish → close

Dependent or overlapping open tickets can remain for a later wave. Failed or unverified useful state remains on its recorded branch and worktree for explicit manual recovery and is never adopted by a later orchestration run.

Beads is the execution tracker for agents. Linear issues passed here resolve to their linked beads tickets, or create them, and Linear is then mirrored for human visibility only.

Quick Task Workflow

For smaller tasks that don't need full design:

/workflow-commands:task "Add logging to auth middleware"
# ... work directly ...
/workflow-commands:verify
git commit -m "feat: add auth middleware logging"

Bug Fix Workflow

/workflow-commands:bug "Users can't login on mobile"
# ... investigate root cause ...
# ... fix it ...
/workflow-commands:verify
git commit -m "fix: mobile login issue"

Resume Work

/workflow-commands:continue beads-a1b2   # Resume specific task
# or
/workflow-commands:continue              # Show available tasks

Customization

Project-Specific Guidance

Run /workflow-commands:project-init --with-guidance to create:

.ed3d/
├── design-plan-guidance.md       # Customize design phase
└── implementation-plan-guidance.md  # Customize implementation

design-plan-guidance.md

# Design Plan Guidance

## Domain Context
<!-- Describe your project's domain, terminology, key concepts -->

## Architectural Constraints
<!-- Decisions that designs must follow -->
- All services must be stateless
- Use PostgreSQL for persistence
- REST APIs only (no GraphQL)

## Technology Stack

### Required
- Python 3.11+
- FastAPI
- PostgreSQL

### Preferred
- Redis for caching
- Celery for async tasks

### Forbidden
- MongoDB (we standardized on PostgreSQL)
- Synchronous external API calls in request handlers

## Stakeholders
<!-- Who approves designs? -->

## Design Conventions
<!-- Project-specific patterns -->

implementation-plan-guidance.md

# Implementation Plan Guidance

## Coding Standards
<!-- Reference style guide or key conventions -->
- Follow PEP 8
- Use type hints everywhere
- Docstrings for public functions

## Testing Requirements
- Minimum 80% coverage for new code
- Integration tests for API endpoints
- Unit tests for business logic

## Review Criteria
- Security review for auth changes
- Performance review for database changes

## Quality Gates
- All tests pass
- No new lint errors
- Type check passes (mypy)
- Pre-commit hooks pass

## Build & Deploy
- CI runs on push to main
- Deploy via GitHub Actions

Verify Customization

The verify skill auto-detects tooling, but you can influence it by:

Adding config files - The skill looks for:
- pytest.ini, pyproject.toml (pytest)
- ruff.toml, .ruff.toml (ruff)
- .pre-commit-config.yaml (pre-commit)
- etc.

Package.json scripts - Define test, lint scripts:

{
  "scripts": {
    "test": "vitest run",
    "lint": "eslint . && prettier --check ."
  }
}

Makefile targets - The skill checks for:

test:
    pytest -v

lint:
    ruff check . && mypy .

verify: test lint
    pre-commit run --all-files

Decision Logging Guidelines

The workflow commands explicitly log decisions. For best results:

Always include:

What was decided
What alternatives were considered
Why this choice was made
Confidence level (50 = tentative, 90 = very confident, scale 0-100)

Good example:

deciduous add decision "Chose Redis for session storage because: 1) sub-ms latency required, 2) built-in TTL for session expiry, 3) team already has Redis expertise. Considered: PostgreSQL (too slow), Memcached (no persistence)." -c 85

Bad example:

deciduous add decision "Using Redis" -c 80

Structure

kyle-claude-plugins/
├── .claude-plugin/
│   └── marketplace.json
├── plugins/
│   ├── workflow-commands/
│   │   ├── .claude-plugin/plugin.json
│   │   ├── commands/
│   │   │   ├── project-init.md
│   │   │   ├── intake.md
│   │   │   ├── explore.md
│   │   │   ├── design.md
│   │   │   ├── plan.md
│   │   │   ├── execute.md
│   │   │   ├── verify.md
│   │   │   ├── fix-pr-review.md
│   │   │   ├── task.md          # NEW
│   │   │   ├── bug.md           # NEW
│   │   │   ├── continue.md
│   │   │   └── orchestrate.md
│   │   └── skills/
│   │       ├── project-init/SKILL.md
│   │       ├── intake/SKILL.md
│   │       ├── exploring/SKILL.md
│   │       ├── designing/SKILL.md
│   │       ├── planning/SKILL.md
│   │       ├── executing/SKILL.md
│   │       ├── verifying/SKILL.md
│   │       ├── pr-review-loop/SKILL.md
│   │       ├── task/SKILL.md              # NEW
│   │       ├── bug/SKILL.md               # NEW
│   │       ├── continue/SKILL.md          # NEW
│   │       ├── orchestrating-beads-tickets/SKILL.md
│   │       └── beads-deciduous-integration/SKILL.md
│   └── tracking-hooks/
│       ├── .claude-plugin/plugin.json
│       └── hooks/
│           ├── hooks.json
│           ├── session-start-tracking.sh
│           └── git-beads-reminder.py
└── README.md

License

MIT

Name		Name	Last commit message	Last commit date
.beads		.beads
.claude-plugin		.claude-plugin
.claude		.claude
.deciduous		.deciduous
.github/workflows		.github/workflows
docs		docs
plugins		plugins
.gitattributes		.gitattributes
.gitignore		.gitignore
AGENTS.md		AGENTS.md
CHANGELOG.md		CHANGELOG.md
CLAUDE.md		CLAUDE.md
README.md		README.md

Latest commit

History

Folders and files

Repository files navigation

Kyle's Claude Code Plugins

Table of Contents

Installation

1. Register the marketplace

2. Install plugins

3. Load the plugins

Updating

Prerequisites

Required Tools

1. beads (Task Tracking)

2. deciduous (Decision Journaling)

3. ed3d-plugins (Core Workflow)

Optional Tools

Linear Integration

GitHub Codex Review

Pre-commit

Commands Reference

/workflow-commands:project-init

/workflow-commands:intake

/workflow-commands:explore

/workflow-commands:design

/workflow-commands:plan

/workflow-commands:execute

/workflow-commands:verify

/workflow-commands:fix-pr-review

/workflow-commands:pollinate

/workflow-commands:pollinate-verify

/workflow-commands:task

/workflow-commands:bug

/workflow-commands:continue

/workflow-commands:orchestrate

/documentation-review:review

Entry Points

Workflow

Full Feature Workflow

Parallel Ticket Workflow

Quick Task Workflow

Bug Fix Workflow

Resume Work

Customization

Project-Specific Guidance

design-plan-guidance.md

implementation-plan-guidance.md

Verify Customization

Decision Logging Guidelines

Structure

License

About

Resources

Stars

Watchers

Forks

Releases

Packages

Contributors

Languages

`/workflow-commands:project-init`

`/workflow-commands:intake`

`/workflow-commands:explore`

`/workflow-commands:design`

`/workflow-commands:plan`

`/workflow-commands:execute`

`/workflow-commands:verify`

`/workflow-commands:fix-pr-review`

`/workflow-commands:pollinate`

`/workflow-commands:pollinate-verify`

`/workflow-commands:task`

`/workflow-commands:bug`

`/workflow-commands:continue`

`/workflow-commands:orchestrate`

`/documentation-review:review`