Name: starry-harness
Author: josephjoshua

Stats

Actions

Available In

Tags

starry-harness

A Claude Code plugin for systematic StarryOS kernel development. Provides bug hunting with Linux comparison testing, kernel internal auditing, performance benchmarking, application compatibility testing, code quality enforcement, upstream submission preparation, and structured reporting — backed by deterministic static analysis tools that eliminate guesswork.

Installation

Add to ~/.claude/settings.json:

{ "extraKnownMarketplaces": { "starry-harness": { "source": { "source": "github", "repo": "JosephJoshua/starry-harness" } } }, "enabledPlugins": { "starry-harness@starry-harness": true } }

Restart Claude Code.

Skills

Skill

What it does

hunt-bugs

Syscall bug discovery: generate test, run on Docker Linux, run on StarryOS, diff, fix. Linux must pass first.

audit-kernel

Deep kernel internal audit beyond syscalls: scheduler, memory, concurrency, signals, filesystem. Uses lockdep, stress testing, property-based tests.

benchmark

Performance measurement against Linux baselines. Profile, optimize, re-measure.

test-app

Run real Linux applications (Nginx, PostgreSQL, Python, etc.) on StarryOS. Strace profiling, gap analysis, blocker fixes.

review-quality

Code quality gate for kernel changes. Rust idioms, safety, API design, framework reuse.

report

Structured bug reports, benchmark reports, app compatibility reports, and a running work journal.

evolve

Autonomous target selection with sweep/deep modes. Picks what to work on based on coverage gaps and past effectiveness. Enforces the review pipeline.

start-submission

Prepares upstream PRs: fresh clone, minimal fix port, test format conversion, verification, Chinese PR draft. Does everything except gh pr create.

check-upstream

Compares known.json against open/merged upstream PRs to identify already-fixed bugs, claimed bugs, and safe-to-submit bugs.

Agents

Agent

Role

linux-comparator

Runs tests on Docker Linux and StarryOS, produces structured diff

kernel-reviewer

Read-only code quality review with fresh context

bug-triager

Classifies bugs by category and severity, recommends fix priority

Deterministic Tools

Static analysis scripts that produce ground-truth output. The agent interprets results — the tools themselves cannot hallucinate.

Script

What it does

abi-check.py

Compares StarryOS syscall arg counts against Linux SYSCALL_DEFINE signatures. Each entry sourced from kernel v6.12 with verifiable URLs.

lock-order-graph.py

Builds a directed graph of lock acquisitions, detects cycles (deadlocks). Rust ownership-aware: distinguishes let guard = x.lock() from x.lock().method().

pattern-scanner.py

Scans kernel source against regex rules (9 default patterns). Rules evolve as new bug classes are found.

kernel-graph.py

Maps all 204 syscalls to subsystems, files, locks, and unsafe blocks.

change-tracker.py

Identifies which tests need re-running based on git diff since last run.

Test Pipeline

Script

What it does

pipeline.sh

Full compile → inject → build → QEMU boot → result capture. Supports --arch riscv64|aarch64|x86_64|loongarch64.

linux-ref-test.sh

Compile and run a C test inside Docker Linux. Supports --arch for cross-arch comparison via QEMU user-mode.

stress-test.sh

Multi-run test execution with SMP sweeping (--smp 1,2,4) and timeout-based deadlock detection.

regression-check.sh

Runs all tests in known.json, compares against expected pass/fail counts, flags regressions.

strace-profiler.sh

Runs an application under strace in Docker, produces a structured syscall profile with gap analysis.

convert-test.py

Converts starry_test.h test format to upstream test_framework.h format.

update-known.sh

Updates known.json with results from a pipeline run.

Utility Scripts

Script

What it does

man-lookup.sh

Fetches syscall man pages (local, Docker, or man7.org).

journal-entry.sh

Appends structured entries to the work journal.

draft-pr.sh

Generates a PR draft markdown file with a ready-to-paste gh pr create command.

How it works

starry-harness

Installation

Add to ~/.claude/settings.json:

{
  "extraKnownMarketplaces": {
    "starry-harness": {
      "source": { "source": "github", "repo": "JosephJoshua/starry-harness" }
    }
  },
  "enabledPlugins": {
    "starry-harness@starry-harness": true
  }
}

Restart Claude Code.

Skills

Skill	What it does
`hunt-bugs`	Syscall bug discovery: generate test, run on Docker Linux, run on StarryOS, diff, fix. Linux must pass first.
`audit-kernel`	Deep kernel internal audit beyond syscalls: scheduler, memory, concurrency, signals, filesystem. Uses lockdep, stress testing, property-based tests.
`benchmark`	Performance measurement against Linux baselines. Profile, optimize, re-measure.
`test-app`	Run real Linux applications (Nginx, PostgreSQL, Python, etc.) on StarryOS. Strace profiling, gap analysis, blocker fixes.
`review-quality`	Code quality gate for kernel changes. Rust idioms, safety, API design, framework reuse.
`report`	Structured bug reports, benchmark reports, app compatibility reports, and a running work journal.
`evolve`	Autonomous target selection with sweep/deep modes. Picks what to work on based on coverage gaps and past effectiveness. Enforces the review pipeline.
`start-submission`	Prepares upstream PRs: fresh clone, minimal fix port, test format conversion, verification, Chinese PR draft. Does everything except `gh pr create`.
`check-upstream`	Compares known.json against open/merged upstream PRs to identify already-fixed bugs, claimed bugs, and safe-to-submit bugs.

Agents

Agent	Role
`linux-comparator`	Runs tests on Docker Linux and StarryOS, produces structured diff
`kernel-reviewer`	Read-only code quality review with fresh context
`bug-triager`	Classifies bugs by category and severity, recommends fix priority

Deterministic Tools

Static analysis scripts that produce ground-truth output. The agent interprets results — the tools themselves cannot hallucinate.

Script	What it does
`abi-check.py`	Compares StarryOS syscall arg counts against Linux `SYSCALL_DEFINE` signatures. Each entry sourced from kernel v6.12 with verifiable URLs.
`lock-order-graph.py`	Builds a directed graph of lock acquisitions, detects cycles (deadlocks). Rust ownership-aware: distinguishes `let guard = x.lock()` from `x.lock().method()`.
`pattern-scanner.py`	Scans kernel source against regex rules (9 default patterns). Rules evolve as new bug classes are found.
`kernel-graph.py`	Maps all 204 syscalls to subsystems, files, locks, and unsafe blocks.
`change-tracker.py`	Identifies which tests need re-running based on `git diff` since last run.

Test Pipeline

Script	What it does
`pipeline.sh`	Full compile → inject → build → QEMU boot → result capture. Supports `--arch riscv64\|aarch64\|x86_64\|loongarch64`.
`linux-ref-test.sh`	Compile and run a C test inside Docker Linux. Supports `--arch` for cross-arch comparison via QEMU user-mode.
`stress-test.sh`	Multi-run test execution with SMP sweeping (`--smp 1,2,4`) and timeout-based deadlock detection.
`regression-check.sh`	Runs all tests in `known.json`, compares against expected pass/fail counts, flags regressions.
`strace-profiler.sh`	Runs an application under strace in Docker, produces a structured syscall profile with gap analysis.
`convert-test.py`	Converts `starry_test.h` test format to upstream `test_framework.h` format.
`update-known.sh`	Updates `known.json` with results from a pipeline run.

Utility Scripts

Script	What it does
`man-lookup.sh`	Fetches syscall man pages (local, Docker, or man7.org).
`journal-entry.sh`	Appends structured entries to the work journal.
`draft-pr.sh`	Generates a PR draft markdown file with a ready-to-paste `gh pr create` command.

starry-harness

Popularity

What's Inside

Confidence

README

starry-harness

Installation

Skills

Agents

Deterministic Tools

Test Pipeline

Utility Scripts

How it works

Similar Plugins

r-skills

unity-dev-toolkit

everything-claude-code

claude-seo

starry-harness

Installation

Skills

Agents

Deterministic Tools

Test Pipeline

Utility Scripts

How it works

Popularity

Health & Quality

Similar Plugins

r-skills

unity-dev-toolkit

everything-claude-code

claude-seo

dotnet-skills

creative-writing