Skip to content
View shihchengwei-lab's full-sized avatar
1 follower · 0 following

Achievements

Achievement: QuickdrawAchievement: Pair Extraordinairex2Achievement: Pull Sharkx2Achievement: YOLO

Achievements

Achievement: QuickdrawAchievement: Pair Extraordinairex2Achievement: Pull Sharkx2Achievement: YOLO

Block or report shihchengwei-lab

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
shihchengwei-lab/README.md

shihchengwei-lab

Side projects for making AI coding agents easier to inspect, resume, and control.

These repos come from real Claude Code / AI-agent workflows: session handoff, context loss, second-opinion review, agent run recording, companion transparency, and authority separation.

Main map

Agent session control

  • claude-code-session-kit — hooks for context alerts, session startup, and handoff validation.
  • coding-guidelines — UserPromptSubmit hook that injects minimal coding rules every turn.

Audit and review

  • cold-eyes-reviewer — diff-centered second-pass review gate for Claude Code.
  • agentcam — local-first flight recorder for AI coding-agent runs.

Companion / transparency work

  • cinder-capture — captures Claude Code Cinder bubble text and reinjects it into context.
  • Buddy_similar — rebuilds a Buddy-like second-opinion companion with explicit data-flow disclosure.
  • BonziClaude — historical reverse-engineering reference for the defunct buddy_react endpoint.
  • HITCON_2026_talk_raw — raw materials for the Buddy / Cinder transparency forensics talk.

Architecture / research prototypes

  • separation-and-audit-alignment — theory notes on pipeline-level authority separation.
  • separation-and-audit-claude-code — Claude Code reference implementation.
  • fourth-path-local-lab — local prototype and benchmark route-validation lab.

Adjacent tools

  • workslop-checker — checks whether AI-written documents are ready to hand off.
  • generate-high-quality-art-image2 — prompt contract for Image 2.0 generation.

Popular repositories Loading

  1. cold-eyes-reviewer cold-eyes-reviewer Public

    Quality-first, diff-centered review gate for Claude Code. Runs as a Stop hook, auto-tunes review depth, runs selected local checks, and turns blocks into agent repair tasks with fresh-review reruns.

    Python 5 1

  2. separation-and-audit-claude-code separation-and-audit-claude-code Public

    Runnable reference implementation of the separation-and-audit-alignment architecture for Claude Code multi-agent projects.

    Python 2

  3. workslop-checker workslop-checker Public

    A tool for checking whether AI-written documents are ready to hand off, ship, publish, or use for decisions. Not an AI detector.

    1

  4. coding-guidelines coding-guidelines Public

    Coding guidelines hook for Claude Code

    Python 1

  5. cinder-capture cinder-capture Public

    Auto-capture Claude Code companion (Cinder) bubble text via Windows UIAutomation + Stop Hook

    PowerShell

  6. claude-code-session-kit claude-code-session-kit Public

    Session lifecycle hooks and CLAUDE.md templates for Claude Code. Context monitoring, automated handoffs, and battle-tested configurations from 92 real sessions.

    Shell