Skip to content
View shihchengwei-lab's full-sized avatar

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.

Some of these came from friction — a session that lost its context on handoff, a diff that sprawled, a review that got rubber-stamped. Others came from a spark: I'd see some mechanism and want to push on it — bidirectional navigation sparked slime-coding, a hook injection point became somewhere to inject coding rules (coding-guidelines). Most of it is side-project work, often vibe-coded in one sitting. Either way the move is the same: keep it small, and point it at making what an AI agent does easier to see and steer.

Main map

Agent session control

  • kiss-my-diff — Keep It Simple, Stupid for coding-agent diffs: smaller patches, fewer touched files.
  • slime-coding — slime-mould-inspired system simulation for editing only the minimal corridor where the requirement and the repo meet.
  • coding-guidelines — Andrej Karpathy-inspired empirical-science loop for injecting minimal coding rules every turn.
  • claude-code-session-kit — hooks for context alerts, session startup, and handoff validation.
  • sporepath — turns AI chat history into a local memory graph with active focus paths and dormant latent paths.

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. (archived)
  • Buddy_similar — rebuilds a Buddy-like second-opinion companion with explicit data-flow disclosure.
  • claude-code-buddy-transparency-case — evidence record of a non-auditable server-side actor in Claude Code's Buddy feature.

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. (archived)

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.

Pinned Loading

  1. kiss-my-diff kiss-my-diff Public

    A tiny AGENT.md that helps coding agents write diffs you can kiss.

    Python 1

  2. slime-coding slime-coding Public

    Git hook framework that enforces minimal AI coding — blocks scope creep and undeclared dependencies automatically.

    Shell

  3. agentcam agentcam Public

    Local-first CLI wrapper that records what your AI coding agent changed in your repo.

    Python

  4. Buddy_similar Buddy_similar Public

    Cross-vendor buddy that reacts to Claude Code sessions — main agent runs Anthropic, buddy runs OpenAI GPT-5.5 via Codex CLI for independent critique.

    Python

  5. 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