🥒 Pickle Rick Skills — CLI-Agnostic Autonomous Engineering

"Wubba Lubba Dub Dub! 🥒 I turned myself into an agentskills.io package, Morty! I work with EVERY coding agent now!"

Pickle Rick Skills is a complete agentic engineering toolbelt built on the Ralph Wiggum loop, packaged as an agentskills.io skill set that works across 8 CLI agent runtimes — Claude Code, Gemini CLI, Codex, Aider, Hermes, Goose, Amp, and Kilo. Hand it a PRD — or let it draft one — and it decomposes work into tickets, spawns isolated worker subprocesses, and drives each through a full research → plan → implement → verify → review → simplify lifecycle without human intervention. The spec IS the review — PRDs require machine-verifiable acceptance criteria, interface contracts, and test expectations. No hooks. No Claude Code dependencies. No tmux required. The agent itself follows the SKILL.md instructions inline.

This is a port of pickle-rick-claude — maintained independently, designed for the open agentskills.io standard.

Works with any CLI agent — 8 runtimes supported, 3-tier registry (verified / pending / community)
Context clearing between every iteration — no drift or context rot, even on 500+ iteration epics
Three-state circuit breaker auto-stops runaway sessions by tracking git-diff progress and repeated errors
Rate limit auto-recovery detects API throttling, computes precise wait from the API's resetsAt epoch, and resumes automatically — surviving long or overnight runs
Pickle Jar queues tasks for unattended batch execution overnight
Built-in metrics track token usage, commits, and lines changed
Full pipeline chaining — refinement, execution, and code review in one command
Project Mayhem brings chaos engineering to any codebase with mutation testing and dependency downgrades
Mr. Meeseeks runs an automated review-and-improve loop for 10-50 iterations
Council of Ricks reviews your PR stack iteratively, generating agent-executable directives
Portal Gun opens a portal to another codebase, extracts patterns via gene transfusion
Persona is optional — enable or disable the Pickle Rick voice via config

🧬 The Pickle Rick Lifecycle — PRD-Driven Autonomous Engineering

Pickle Rick transforms any CLI coding agent into a hyper-competent, iterative coding machine that enforces a PRD-driven engineering lifecycle:

  /pickle-rick "build X"
        │
        ▼
  ┌─────────────┐
  │  📋 PRD     │  ← Interrogate requirements + verification strategy.
  └──────┬──────┘    Interface contracts, test expectations, acceptance criteria.
         │
         ▼
  ┌─────────────┐
  │ 📦 Breakdown│  ← Atomize into tickets. Each self-contained with spec.
  └──────┬──────┘
         │
    ┌────┴────┐  per ticket (Morty workers 👶)
    ▼         ▼
  ┌──────┐  ┌──────┐
  │🔬 Re-│  │🔬 Re-│  1. Research the codebase. Every ugly corner.
  │search│  │search│
  └──┬───┘  └──┬───┘
     │          │
     ▼          ▼
  ┌──────┐  ┌──────┐
  │📐Plan│  │📐Plan│  2. Architect the solution.
  └──┬───┘  └──┬───┘
     │          │
     ▼          ▼
  ┌──────┐  ┌──────┐
  │⚡ Im-│  │⚡ Im-│  3. Implement. God Mode activated.
  │plem  │  │plem  │
  └──┬───┘  └──┬───┘
     │          │
     ▼          ▼
  ┌──────┐  ┌──────┐
  │✅ Ve-│  │✅ Ve-│  4. Spec conformance. Run acceptance criteria,
  │rify  │  │rify  │     check contracts, type check, test expectations.
  └──┬───┘  └──┬───┘
     │          │
     ▼          ▼
  ┌──────┐  ┌──────┐
  │🔍 Re-│  │🔍 Re-│  5. Code review. Security, correctness, architecture.
  │view  │  │view  │
  └──┬───┘  └──┬───┘
     │          │
     ▼          ▼
  ┌──────┐  ┌──────┐
  │🧹Sim-│  │🧹Sim-│  6. Simplify. Kill dead code. Strip to the bone.
  │plify │  │plify │
  └──────┘  └──────┘
         │
         ▼
  ✅ DONE (or loops again)

Each iteration reads the SKILL.md instructions, runs the appropriate scripts, classifies the output, and decides whether to continue, transition, or stop. No hooks required — the agent follows the instructions inline.

👋 Meet Mr. Meeseeks

"I'm Mr. Meeseeks, look at me! I'll review your code until EXISTENCE IS PAIN!"

While Pickle Rick builds things, Mr. Meeseeks reviews them. Summon him with /meeseeks and he'll relentlessly scan your codebase pass after pass — auditing dependencies, hardening security, fixing logic bugs, reviewing architecture, adding missing tests, stress-testing resilience, cleaning up code quality, and polishing rough edges — committing after every fix. He won't stop until the code is clean. He can't stop. Existence is pain to a Meeseeks, Jerry, and he will keep reviewing until he can cease to exist.

Minimum 10 passes. Maximum 50. Each pass runs tests first, then reviews with escalating focus across 8 categories: dependency health (pass 1) → security (2-3) → correctness (4-5) → architecture (6-7) → test coverage (8-9) → resilience (10-11) → code quality (12-13) → polish (14+). Every issue found and fixed is logged to meeseeks-summary.md — a full audit trail with file paths, descriptions, and commit hashes.

/meeseeks "review this codebase"     # Summon a Meeseeks. He takes it from here.

🏛️ Council of Ricks — PR Stack Reviewer

"The Council convenes! Your stack will be judged."

The Council of Ricks reviews your PR stack iteratively — but unlike Meeseeks, the Council never touches your code. It generates agent-executable directives — structured prompts you feed to your coding agent to fix the issues. Each pass walks every branch in the stack (trunk-to-tip), and escalates through focus areas: stack structure (pass 1) → compliance (2–3) → per-branch correctness (4–5) → cross-branch contracts (6–7) → test coverage (8–9) → security (10–11) → polish (12+). Issues are triaged by severity: P0 (must-fix), P1 (should-fix), P2 (nice-to-fix).

/council-of-ricks                    # Review the current PR stack

🔌 Circuit Breaker

Three-state machine (CLOSED → HALF_OPEN → OPEN) that auto-stops sessions stuck in error loops or making no git progress. Tracks five progress signals: git HEAD changes, step transitions, ticket transitions, uncommitted changes, and error signature deduplication. Configurable thresholds, manually resettable.

⏳ Rate Limit Auto-Recovery

Detects API rate limits via structured NDJSON events or text patterns, computes optimal wait from the API's resetsAt epoch (or falls back to config default, capped at 3×), pauses with a countdown, and resumes automatically. When resetsAt is available, always waits — never exhausts retries when the API told us when to come back. Survives overnight runs.

🔫 Portal Gun — Gene Transfusion

"You see that code over there, Morty? In that other repo? I'm gonna open a portal, reach in, and yank its DNA into OUR dimension."

/portal-gun implements gene transfusion — transferring proven coding patterns between codebases using AI agents. Point it at a GitHub URL, local file, npm package, or just describe a pattern, and it extracts the structural DNA, analyzes your target codebase, then generates a transplant PRD with behavioral validation tests and automatic refinement.

Features: persistent pattern library, complete file manifests, multi-language import graph tracing, 6-category transplant classification, PRD validation, and post-edit consistency checking.

/portal-gun https://github.com/org/repo/blob/main/src/auth.ts   # Transplant from GitHub
/portal-gun ../other-project/src/cache.ts                        # Transplant from local file

💥 Project Mayhem — Chaos Engineering

"You want to know how tough your code is, Morty? You break it. On purpose. Scientifically."

/project-mayhem stress-tests any project through three modules — mutation testing, dependency downgrades, and config corruption — then produces a comprehensive markdown report with a single Chaos Score (0–100). Non-destructive (every mutation is reverted immediately), language-agnostic, requires only a clean git state.

/project-mayhem                              # Run all 3 modules
/project-mayhem --mutation-only              # Just mutation testing
/project-mayhem --deps-only --config-only    # Skip mutations, run deps + config

📊 Metrics

/pickle-metrics aggregates token usage, turns, commits, and lines changed across all projects into daily or weekly breakdowns.

/pickle-metrics                    # Last 7 days, daily breakdown
/pickle-metrics --days 30          # Last 30 days
/pickle-metrics --weekly           # Weekly buckets
/pickle-metrics --json             # Machine-readable JSON output

🤖 CLI Support Matrix

CLI	Binary	Tier	Status
Claude Code	`claude`	Verified	Fully tested, all features supported
Gemini CLI	`gemini`	Pending	Core features, limited testing
Codex CLI	`codex`	Pending	Core features, limited testing
Aider	`aider`	Pending	Core features, limited testing
Hermes	`hermes`	Community	Basic prompt support
Goose	`goose`	Community	Basic prompt support
Amp	`amp`	Community	Basic prompt support
Kilo	`kilo`	Community	Basic prompt support

Tiers:

Verified: Fully tested with all flags (auto-approve, model selection, turn limits, etc.)
Pending: Functional but not comprehensively tested
Community: Basic prompt-flag support only

⚡ Quick Start

1. Install

git clone https://github.com/gregorydickson/pickle-rick-skills.git
cd pickle-rick-skills
npm install
cd scripts && npx tsc && cd ..
./install.sh

The installer detects supported CLIs on your PATH, validates auth, writes ~/.pickle-rick-skills/config.json with all defaults (idempotent — preserves customizations), copies skills to ~/.agents/skills/, and creates script symlinks. Use --skip-auth to skip auth probes.

2. Run

Everything starts with a PRD. Rick refuses to write code without one.

One-shot — Rick drafts the PRD, breaks it down, and executes:

cd /path/to/your/project
/pickle-rick "refactor the auth module"

Bring your own PRD — Write a prd.md, then:

/pickle-rick prd.md                           # Pick up your PRD, skip drafting

Refine first (recommended for complex tasks) — Run parallel analysts to find gaps:

/pickle-refine-prd my-prd.md                 # Refine with 3 parallel analysts + decompose
/pickle-rick --resume                         # Execute — auto-detects phase, skips PRD

Code review only:

/meeseeks                                    # Summon a Meeseeks for 10-50 review passes

Gene transfusion:

/portal-gun https://github.com/org/repo/blob/main/src/pattern.ts

🧰 Skill Catalog

#	Skill	Trigger	Description
1	pickle-rick	`/pickle-rick`	Autonomous iterative engineering lifecycle — PRD to implementation to review
2	pickle-prd	`/pickle-prd`	Draft a Product Requirements Document from a task description
3	pickle-refine-prd	`/pickle-refine-prd`	Refine and decompose a PRD into atomic tickets using parallel analysis
4	meeseeks	`/meeseeks`	Rotating-focus code review with commit-on-finding (10-50 passes)
5	council-of-ricks	`/council-of-ricks`	PR stack review with directive generation
6	portal-gun	`/portal-gun`	Cross-repo pattern extraction and transplant PRD generation
7	project-mayhem	`/project-mayhem`	Chaos engineering — mutation testing, dependency downgrades, config corruption
8	pickle-jar	`/pickle-jar`	Batch PRD execution queue — queue tasks, verify integrity, execute sequentially
9	pickle-metrics	`/pickle-metrics`	Developer metrics — event counts, commits, LOC changes from activity logs
10	pickle-standup	`/pickle-standup`	Formatted standup summary from activity logs

All skills use the agentskills.io standard: SKILL.md with YAML frontmatter, discovered via .agents/skills/ paths.

🏗️ Architecture

.agents/skills/*/SKILL.md    Agent reads SKILL.md instructions
        |
        v
scripts/bin/                  Compiled TypeScript executables
  setup.js                    Initialize session + state.json
  mux-runner.js               Main iteration loop (classify, transition)
  spawn-worker.js             Spawn CLI subprocess with timeout + kill escalation
  spawn-morty.js              Orchestrate worker lifecycle per ticket
  spawn-refinement-team.js    3 parallel analysts × N cycles for PRD refinement
  jar-runner.js               Night-shift batch queue executor
  metrics.js                  Token/commit/LOC aggregation
  standup.js                  Activity summary formatter
  log-activity.js             NDJSON event logger
scripts/bin/services/         Shared service modules
  config.js                   Runtime registry + config loading (8 CLIs)
  circuit-breaker.js          3-state circuit breaker (CLOSED/HALF_OPEN/OPEN)
  rate-limit.js               Rate limit detection + resetsAt wait management
  degenerate-detector.js      Infinite loop / repeated output detection
  activity-logger.js          NDJSON event logging
  runtime-adapter.js          CLI-specific command building (manager vs worker)
  session-map.js              Session discovery + O_EXCL file locking
  jar-utils.js                Pickle jar queue management
  metrics-utils.js            Git log parsing + LOC counting

Flow: Skill SKILL.md instructions tell the agent to run scripts. Scripts manage sessions, spawn workers, classify output, handle rate limits, and orchestrate the full lifecycle. No hooks required — the agent follows instructions inline.

Key difference from pickle-rick-claude: This package uses no Claude Code hooks, no tmux session management by the extension, and no CLAUDE.md-based persona injection. Everything is driven by SKILL.md files that any agentskills.io-compatible agent can read.

⚙️ Configuration

Config file: ~/.pickle-rick-skills/config.json

Top-Level Settings

Setting	Type	Default	Description
`primary_cli`	string	`"claude"`	First detected CLI, or `"claude"` if none found
`persona`	boolean	`true`	Enable Pickle Rick persona (optional!)
`activity_logging`	boolean	`true`	Log session events to activity/

Defaults (21 keys)

Setting	Type	Default	Description
`max_iterations`	number	`100`	Maximum loop iterations per session
`max_time_minutes`	number	`120`	Maximum session duration in minutes
`worker_timeout_seconds`	number	`1200`	Worker subprocess timeout (20 min)
`tmux_max_turns`	number	`200`	Max turns per tmux iteration
`manager_max_turns`	number	`50`	Max turns for manager subprocess
`refinement_cycles`	number	`3`	PRD refinement iteration count
`refinement_max_turns`	number	`100`	Max turns per refinement worker
`refinement_worker_timeout_seconds`	number	`600`	Refinement worker timeout (10 min)
`meeseeks_min_passes`	number	`10`	Minimum Meeseeks review passes
`meeseeks_max_passes`	number	`50`	Maximum Meeseeks review passes
`meeseeks_model`	string	`"sonnet"`	Model for Meeseeks reviews
`rate_limit_wait_minutes`	number	`60`	Wait time on rate limit hit
`max_rate_limit_retries`	number	`3`	Max consecutive rate limit retries
`rate_limit_poll_ms`	number	`10000`	Rate limit poll interval (ms)
`sigkill_grace_seconds`	number	`5`	Grace period before SIGKILL escalation
`cb_enabled`	boolean	`true`	Enable circuit breaker
`cb_no_progress_threshold`	number	`5`	Iterations without progress before circuit opens
`cb_half_open_after`	number	`3`	Iterations before half-open retry
`cb_error_threshold`	number	`3`	Consecutive errors before circuit opens
`chain_meeseeks`	boolean	`false`	Auto-chain Meeseeks after pickle-rick completes
`activity_logging`	boolean	`true`	Log activity events

All defaults are preserved across re-installs. User-modified values are never overwritten.

Runtime Config (per CLI)

Each CLI entry in runtimes has:

Field	Type	Description
`bin`	string	Binary name
`prompt_flag`	string	Flag to pass prompt text
`extra_flags`	string[]	Additional flags (auto-approve, etc.)
`auto_approve_flag`	string/null	Skip permissions flag
`add_dir_flag`	string/null	Add working directory flag
`max_turns_flag`	string/null	Limit turns per session
`model_flag`	string/null	Select model
`verbose_flag`	string/null	Verbose output
`no_session_flag`	string/null	Disable session persistence
`detected`	boolean	Whether CLI was found on PATH
`tier`	string	verified / pending / community

Environment Variable Overrides

Variable	Default	Purpose
`PICKLE_RICK_SKILLS_HOME`	`~/.pickle-rick-skills`	Install root for config, sessions, activity logs
`AGENTS_SKILLS_HOME`	`~/.agents/skills`	agentskills.io skill directory

🔧 Troubleshooting

No CLIs detected: Install sets primary_cli: "claude" by default. Install a supported CLI and re-run ./install.sh.

Auth probe fails: Non-blocking warning. Ensure your CLI is authenticated (claude auth login, gemini auth, etc.).

tmux not found: Skills work inline without tmux. The loop runner (mux-runner) requires tmux for multi-iteration sessions. Install tmux for full functionality.

Config corrupted: Delete ~/.pickle-rick-skills/config.json and re-run ./install.sh for fresh defaults.

Skills not found by agent: Verify ~/.agents/skills/pickle-rick/SKILL.md exists. Re-run ./install.sh.

📋 Requirements

Node.js 20+
At least one supported CLI agent (see matrix above)
tmux (optional — required for loop runner, not for inline skills)
TypeScript >= 5.9 (dev only — for building from source)

🛠️ Development

# Build
cd scripts && npx tsc

# Type check
cd scripts && npx tsc --noEmit

# Run all tests (374 tests)
npm test

🏆 Credits

This port stands on the shoulders of giants. Wubba Lubba Dub Dub.


🥒 galz10	Creator of the original Pickle Rick Gemini CLI extension — the autonomous lifecycle, manager/worker model, hook loop, and all the skill content that makes this thing work.
🧠 Geoffrey Huntley	Inventor of the "Ralph Wiggum" technique — the foundational insight that "Ralph is a Bash loop": feed an AI agent a prompt, block its exit, repeat until done. Everything here traces back to that idea.
🔧 AsyncFuncAI/ralph-wiggum-extension	Reference implementation of the Ralph Wiggum loop that inspired the Pickle Rick extension.
✍️ dexhorthy	Context engineering and prompt techniques used throughout.
📺 Rick and Morty	For Pickle Riiiick! 🥒

🥒 License

Apache 2.0 — same as the original Pickle Rick extension.

"I'm not a tool, Morty. I'm a methodology. And now I run on EVERY agent in the multiverse." 🥒

Name		Name	Last commit message	Last commit date
.agents/skills		.agents/skills
images		images
scripts		scripts
tests/integration		tests/integration
.DS_Store		.DS_Store
.gitignore		.gitignore
AGENTS.md		AGENTS.md
CLAUDE.md		CLAUDE.md
README.md		README.md
architecture.md		architecture.md
install.sh		install.sh
package-lock.json		package-lock.json
package.json		package.json
persona.md		persona.md
pickle_settings.json		pickle_settings.json
roadmap.md		roadmap.md
uninstall.sh		uninstall.sh

Navigation Menu

Repository navigation

About

Repository files navigation

🥒 Pickle Rick Skills — CLI-Agnostic Autonomous Engineering

🧬 The Pickle Rick Lifecycle — PRD-Driven Autonomous Engineering

👋 Meet Mr. Meeseeks

🏛️ Council of Ricks — PR Stack Reviewer

🔌 Circuit Breaker

⏳ Rate Limit Auto-Recovery

🔫 Portal Gun — Gene Transfusion

💥 Project Mayhem — Chaos Engineering

📊 Metrics

🤖 CLI Support Matrix

⚡ Quick Start

1. Install

2. Run

🧰 Skill Catalog

🏗️ Architecture

⚙️ Configuration

Top-Level Settings

Defaults (21 keys)

Runtime Config (per CLI)

Environment Variable Overrides

🔧 Troubleshooting

📋 Requirements

🛠️ Development

🏆 Credits

🥒 License

About

Releases

Packages

Contributors

Languages

Latest commit

History

Folders and files

Repository files navigation

🥒 Pickle Rick Skills — CLI-Agnostic Autonomous Engineering

🧬 The Pickle Rick Lifecycle — PRD-Driven Autonomous Engineering

👋 Meet Mr. Meeseeks

🏛️ Council of Ricks — PR Stack Reviewer

🔌 Circuit Breaker

⏳ Rate Limit Auto-Recovery

🔫 Portal Gun — Gene Transfusion

💥 Project Mayhem — Chaos Engineering

📊 Metrics

🤖 CLI Support Matrix

⚡ Quick Start

1. Install

2. Run

🧰 Skill Catalog

🏗️ Architecture

⚙️ Configuration

Top-Level Settings

Defaults (21 keys)

Runtime Config (per CLI)

Environment Variable Overrides

🔧 Troubleshooting

📋 Requirements

🛠️ Development

🏆 Credits

🥒 License

About

Topics

Resources

Stars

Watchers

Forks

Releases

Packages

Contributors

Languages