Galeharness Cli

Compounding Engineering Plugin

AI-powered development tools that get smarter with every use. Make each unit of engineering work easier than the last.

Getting Started

After installing, run /gh:setup in any project. It diagnoses your environment, installs missing tools, and bootstraps project config in one interactive flow.

HKTMemory Public Knowledge Root

Gale-managed memory calls use gale-memory as the runtime boundary. By default, task memory resolves to:

~/.galeharness/knowledge/<project>/hkt-memory

Run gale-memory status to see the active root, source, and migration state. status only diagnoses; it does not copy legacy files.

Existing project-local memory/ directories are copied into the public root by gale-memory migrate or task runtime on first migrating use, but they are not deleted or renamed. The copy-first migration preserves Markdown memories and the migration manifest while skipping derived caches such as *.db, vector stores, session transcript indexes, and _lifecycle/events.jsonl.

Override order: gale-memory --memory-dir, existing HKT_MEMORY_DIR, memory.hkt_memory_dir in ~/.galeharness/config.{json,yaml}, then the derived public root. Bare hkt-memory keeps its upstream default path behavior; use gale-memory or set HKT_MEMORY_DIR when you want the public knowledge root.

Document Language

Control the output language for workflow skills via .compound-engineering/config.local.yaml:

language: zh-CN  # zh-CN (Chinese, default) or en (English)

Affected skills: gh:brainstorm, gh:plan, gh:compound, gh:compound-refresh, gh:ideate, document-review

What gets translated:

• Prose content: paragraphs, list items, table content
• Finding descriptions in document-review (including headless mode JSON)

What stays in English:

• Section headers (## Problem Frame, ## Requirements, etc.)
• YAML frontmatter keys (title, date, category, etc.)
• Code blocks, inline code, file paths, URLs

Default: zh-CN when not configured. To opt out, set language: en.

Components

Component	Count
Agents	50+
Skills	42

Recent upstream-sync highlights

The latest upstream capability sync preserves GaleHarness-specific gh:* workflows while selectively importing high-value fixes from the reference compound-engineering plugin. This was a targeted sync, not a wholesale baseline bump, so local skills such as /gh:nexus remain intact.

Highlights:

• resolve-pr-feedback now paginates GitHub GraphQL review threads so long-lived PRs do not silently miss comments.
• web-researcher no longer assumes Claude-only WebSearch / WebFetch tools, making research guidance safer across runtimes.
• coherence-reviewer has stronger document-type adaptation and safer safe_auto consistency rules for review-driven fixes.
• git-worktree resolves helper scripts through ${CLAUDE_SKILL_DIR:-.}, improving both marketplace and local plugin installs.
• gh-polish-beta project detection is compatible with macOS Bash 3.2.
• CLI install / convert / sync detection now respects $CODEX_HOME, keeping Codex profiles isolated from the default home.

Validation covered the full Bun test suite, release metadata validation, namespace leak checks, and Linux + Windows CI.

Skills

Core Workflow

The primary entries form an end-to-end engineering chain:

/gh:ideate -> /gh:brainstorm -> /gh:plan -> /gh:work or /gh:work-beta or /gh:work-x -> /gh:review -> /gh-demo-reel -> /gh:compound

Use the horizontal skills (/gh:debug, /gh:optimize, /gh:simplify-code, /gh:product-pulse, /gh:sessions, /gh:slack-research) when the situation calls for them rather than forcing every task through the full chain.

Skill	When to use it	Output / value
/gh:ideate	You want grounded improvement ideas or alternative directions before committing to a requirement.	Ranked ideas with constraints, risks, and next-step candidates.
/gh:brainstorm	A request is still fuzzy and needs problem framing, assumptions, non-goals, and success criteria.	A requirement-quality brief that can feed /gh:plan.
/gh:plan	A requirement is ready to decompose into implementation steps or an existing plan needs deepening.	A structured plan with files, risks, tests, and confidence checks.
/gh:work	Execute a scoped development task or plan with normal quality gates.	Implemented code plus verification evidence.
/gh:work-beta	Execute work while conserving main-agent context via external delegation/Codex-assisted split work.	Same target as /gh:work, with beta delegation tradeoffs.
/gh:work-x	Execute iOS Swift/ObjC work where Morph-X blueprint constraints and similarity audit reduce template-code repetition risk.	iOS-focused implementation with Morph-X transform/audit evidence.
/gh:debug	Investigate errors, test failures, stack traces, or production symptoms where root cause is unknown.	Causal diagnosis, testable hypothesis, fix, and regression verification.
/gh:debug-x	Debug iOS Swift/ObjC issues under Morph-X blueprint constraints.	iOS root-cause fix plus similarity/safety audit.
/gh:review	Run independent pre-PR or PR review with tiered personas and confidence filtering.	Deduplicated findings, severity, confidence, and fix recommendations.
/gh-demo-reel	Capture visual or terminal evidence after a change is working and before/while preparing PR evidence.	GIF, screenshot, or terminal recording URL/path suitable for PR notes.
/gh:compound	A problem was solved and should become reusable team knowledge.	A durable learning/solution document for future retrieval.
/gh:compound-refresh	Existing docs/solutions/ knowledge may be stale, duplicated, or drifting.	Keep/update/replace/archive decisions and refreshed docs.
/gh:nexus	Analyze repos with GitNexus code intelligence (symbols, cypher queries, impact analysis).	Indexed repo insights, symbol context, change-impact reports.

For /gh:optimize, see skills/gh-optimize/README.md for usage guidance, example specs, and links to the schema and workflow docs.

Strategy, product, and quality helpers

Skill	When to use it	Output / value
/gh:strategy	A repo lacks a crisp product anchor, or the team needs to realign on problem, users, metrics, and tracks.	Creates or updates root STRATEGY.md; prefer before large multi-track work.
/gh:product-pulse	You need a time-windowed read-only pulse of user experience, errors, performance, and quality signals.	Product pulse report with risks, regressions, and follow-up candidates.
/gh:simplify-code	Recent changes work but feel too complex, duplicated, or hard to reuse.	Behavior-preserving simplification and clarity improvements.
/gh:optimize	There is a measurable target such as performance, search quality, prompt quality, ranking, or clustering.	Metric-driven optimization loop with parallel experiments and judge/measurement gates.
/gh:polish-beta	A PR or feature needs browser/dev-server polish after review and CI, especially UI/interaction cleanup.	Testable polish checklist and scoped fix dispatches; beta surface.

Workflow Guardrails

The core gh: workflow includes Karpathy-inspired guardrails without requiring a separate skill. gh:brainstorm challenges problem framing and separates assumptions, non-goals, and success criteria; gh:plan requires complexity to trace to requirements, risks, or constraints; gh:work keeps non-trivial execution tied to a minimal change contract and surgical diffs; gh:review checks diff hygiene against intent and plan scope.

For the source references and integration rationale, see docs/brainstorms/2026-04-24-karpathy-guidelines-integration-requirements.md and docs/solutions/workflow-issues/karpathy-guidelines-workflow-guardrails-2026-04-24.md in this repository.

Research & Context

Skill	When to use it	Output / value
/gh:sessions	Ask questions about prior Claude Code, Codex, and Cursor coding-agent sessions for the current repo or topic.	Synthesized history, prior decisions, failures, and relevant evidence.
/gh-session-inventory	Internal/support skill for discovering session files and metadata before choosing sessions to inspect.	JSONL inventory with platform, branch/CWD hints, timestamps, and keyword match counts.
/gh-session-extract	Internal/support skill for extracting a confirmed relevant session's skeleton or error signals.	Bounded skeleton/error excerpts that avoid loading huge raw transcripts.
/gh:slack-research	Search Slack for interpreted organizational context -- decisions, constraints, and discussion arcs.	Digest of relevant messages with synthesized context, not a raw message dump.

Git Workflow

Skill	Description
git-clean-gone-branches	Clean up local branches whose remote tracking branch is gone
git-commit	Create a git commit with a value-communicating message
git-commit-push-pr	Commit, push, and open a PR with an adaptive description; also update an existing PR description, or generate a description on its own without committing
git-worktree	Manage Git worktrees for parallel development

Upstream Sync Workflow

GaleHarnessCLI maintains a local per-commit upstream sync workflow for bringing changes over from the reference upstream repo without collapsing everything into one giant diff.

Repository entry points:

bash scripts/upstream-sync/generate-batch.sh --upstream-repo /path/to/upstream-checkout
bash scripts/upstream-sync/apply-patch-to-worktree.sh .context/galeharness-cli/upstream-sync/<date>/adapted/<patch>.patch

What this workflow records:

• commit-range.txt records the batch start commit, end commit, and next_baseline_candidate
• README.md in the batch directory records the patch table and recommended worktree flow
• .upstream-ref remains the durable baseline and should be updated manually to the batch next_baseline_candidate only after the whole batch has landed

Recommended operating model:

• Generate one dated batch from .upstream-ref -> upstream HEAD
• Process one adapted patch per isolated worktree
• Open one PR per adapted patch
• After all PRs land, write the batch end_commit / next_baseline_candidate into .upstream-ref

Workflow Utilities

Skill	When to use it	Output / value
/report-bug-ce	Report a bug in the compound-engineering plugin.	Structured bug report for plugin maintainers.
/resolve-pr-feedback	Triage and resolve PR review feedback in parallel.	Validated fixes or documented rejected feedback.
/sync	Sync Claude Code config across machines.	Updated local config state.
/test-browser	Run browser tests on pages affected by a PR or branch.	Browser-test evidence and failures.
/test-xcode	Build and test iOS apps on simulator using XcodeBuildMCP.	Xcode build/test evidence.
/gh:setup	Diagnose environment, install missing tools, and bootstrap project config.	Environment report and setup fixes.
/gh:update	Check compound-engineering plugin version and fix stale cache (Claude Code only).	Plugin update/cache status and remediation.

Development Frameworks

Skill	Description
agent-native-architecture	Build AI agents using prompt-native architecture
dhh-rails-style	Write Ruby/Rails code in DHH's 37signals style
frontend-design	Create production-grade frontend interfaces

Review & Quality

Skill	Description
document-review	Review documents using parallel persona agents for role-specific feedback

Content & Collaboration

Skill	Description
proof	Create, edit, and share documents via Proof collaborative editor

Automation & Tools

Skill	Description
gemini-imagegen	Generate and edit images using Google's Gemini API

Beta / Experimental

Skill	Description
gh:polish-beta	Human-in-the-loop polish phase after /gh:review — verifies review + CI, starts a dev server from .claude/launch.json, generates a testable checklist, and dispatches polish sub-agents for fixes. Emits stacked-PR seeds for oversized work
/lfg	Full autonomous engineering workflow

Morph-X CLI Utilities

gale-harness audit --similarity <project> scans Swift/ObjC source against configured baselines and reports statement, token, structural, and control-flow similarity risk. gale-harness morph --apply <project> selects a deterministic Morph-X blueprint/strategy and calls an optional SwiftSyntax adapter when configured. These tools are for reducing template-code repetition risk; they do not guarantee Apple App Review outcomes.

Agents

Agents are specialized subagents invoked by skills — you typically don't call these directly.

Review

Agent	Description
agent-native-reviewer	Verify features are agent-native (action + context parity)
api-contract-reviewer	Detect breaking API contract changes
architecture-strategist	Analyze architectural decisions and compliance
code-simplicity-reviewer	Final pass for simplicity and minimalism
correctness-reviewer	Logic errors, edge cases, state bugs
data-integrity-guardian	Database migrations and data integrity
data-migration-expert	Validate ID mappings match production, check for swapped values
data-migrations-reviewer	Migration safety with confidence calibration
deployment-verification-agent	Create Go/No-Go deployment checklists for risky data changes
dhh-rails-reviewer	Rails review from DHH's perspective
julik-frontend-races-reviewer	Review JavaScript/Stimulus code for race conditions
gale-rails-reviewer	Rails code review with strict conventions
gale-python-reviewer	Python code review with strict conventions
gale-typescript-reviewer	TypeScript code review with strict conventions
maintainability-reviewer	Coupling, complexity, naming, dead code
pattern-recognition-specialist	Analyze code for patterns and anti-patterns
performance-oracle	Performance analysis and optimization
performance-reviewer	Runtime performance with confidence calibration
reliability-reviewer	Production reliability and failure modes
schema-drift-detector	Detect unrelated schema.rb changes in PRs
security-reviewer	Exploitable vulnerabilities with confidence calibration
security-sentinel	Security audits and vulnerability assessments
swift-ios-reviewer	Swift and iOS code review -- SwiftUI state, retain cycles, concurrency, Core Data threading, accessibility
testing-reviewer	Test coverage gaps, weak assertions
project-standards-reviewer	CLAUDE.md and AGENTS.md compliance
adversarial-reviewer	Construct failure scenarios to break implementations across component boundaries

Document Review

Agent	Description
coherence-reviewer	Review documents for internal consistency, contradictions, and terminology drift
design-lens-reviewer	Review plans for missing design decisions, interaction states, and AI slop risk
feasibility-reviewer	Evaluate whether proposed technical approaches will survive contact with reality
product-lens-reviewer	Challenge problem framing, evaluate scope decisions, surface goal misalignment
scope-guardian-reviewer	Challenge unjustified complexity, scope creep, and premature abstractions
security-lens-reviewer	Evaluate plans for security gaps at the plan level (auth, data, APIs)
adversarial-document-reviewer	Challenge premises, surface unstated assumptions, and stress-test decisions

Research

Agent	Description
best-practices-researcher	Gather external best practices and examples
framework-docs-researcher	Research framework documentation and best practices
git-history-analyzer	Analyze git history and code evolution
issue-intelligence-analyst	Analyze GitHub issues to surface recurring themes and pain patterns
learnings-researcher	Search institutional learnings for relevant past solutions
repo-research-analyst	Research repository structure and conventions
session-historian	Search prior Claude Code, Codex, and Cursor sessions for related investigation context
slack-researcher	Search Slack for organizational context relevant to the current task
web-researcher	Perform structured external web research for prior art, market signals, and cross-domain analogies

Design

Agent	Description
design-implementation-reviewer	Verify UI implementations match Figma designs
design-iterator	Iteratively refine UI through systematic design iterations
figma-design-sync	Synchronize web implementations with Figma designs

Workflow

Agent	Description
pr-comment-resolver	Address PR comments and implement fixes
spec-flow-analyzer	Analyze user flows and identify gaps in specifications

Docs

Agent	Description
ankane-readme-writer	Create READMEs following Ankane-style template for Ruby gems

Installation

claude /plugin install compound-engineering

Then run /gh:setup to check your environment and install recommended tools.

Version History

See the repo root CHANGELOG.md for canonical release history.

License

MIT