llm-wb 0.1.0-beta.0

package/README.md

@@ -0,0 +1,98 @@
+# llm-workbench
+
+Turns AI coding conversations into isolated, reproducible Git workspaces.
+
+Run AI coding sessions in isolated git worktrees without dirtying `main`.
+
+## Why use it?
+
+AI coding tools are useful, but they make it easy to lose context, mix unrelated changes, or pollute your main branch.
+
+`llm-workbench` gives every chat session its own branch, worktree, and session log so you can experiment safely and merge only when ready.
+
+### Before LLM-Workbench
+* one long AI conversation trying to solve multiple problems
+* random commits
+* forgot why code changed
+* dirty main
+* abandoned branches
+
+### After LLM-Workbench
+
+```text
+Main
+    │.                         │
+    ▼.                         ▼
+Start chat #1.            Start chat #n
+    │.                         │
+    ▼.                         ▼
+Create worktree/branch.       ...
+    │
+    ▼
+Code with AI
+    │
+    ▼
+Checkpoint
+    │
+    ▼
+Continue tomorrow
+    │
+    ▼
+Merge back to main
+```
+
+* one branch per AI session
+* session logs
+* isolated worktrees
+* merge when ready
+
+## Quick start
+
+Public npm install is the intended user path:
+
+```bash
+cd /path/to/your/repo
+
+npx llm-wb init --dry-run
+npx llm-wb init
+```
+
+Then start and inspect work from that repo:
+
+```bash
+npx llm-wb new "your prompt here"
+npx llm-wb sessions list
+npx llm-wb commit -m "Describe the completed work"
+npx llm-wb merge-main
+```
+
+`llm-wb list` lists installed workbench commands. To list active chat sessions,
+use `llm-wb sessions list`.
+
+This installs into your target repo:
+
+* Adds the `llm-wb` CLI
+* Configures AGENTS.md
+* Installs workbench scripts
+* Enables safe uninstall
+* Creates session logs that preserve decisions, prompts and commit history.
+
+Each new llm-chat session then creates:
+
+- a chat-owned branch
+- a git worktree for that branch
+- a session log/checkpoint record
+- merge-readiness checks before promotion back to `main`
+
+Works with existing repositories. No migration required. Uninstall removes only files owned by LLM Workbench.
+
+## Who is this for?
+
+Developers using Claude Code, Codex, Cursor or Copilot coding assistants who want safer multi-session repo workflows.
+
+## Security Boundaries
+
+* does not push to remote
+* does not rewrite history
+* does not delete branches without approval
+* does not require a specific LLM provider

package/SECURITY.md

@@ -0,0 +1,44 @@
+# Security Policy
+
+`llm-workbench` is local developer tooling. It works with Git repositories,
+shell scripts, worktrees, and generated session logs, so security reports are
+especially important when they involve command execution, file writes, path
+handling, or accidental disclosure of private repo data.
+
+## Reporting A Vulnerability
+
+Please do not open a public issue for a suspected vulnerability.
+
+Report security concerns privately to the maintainer instead. Include:
+
+- the affected command, script, workflow, or install path
+- the operating system and shell, if relevant
+- steps to reproduce
+- whether the issue can write files, execute commands, expose private data, or
+  bypass an approval boundary
+
+If GitHub private vulnerability reporting is enabled for this repo, use that.
+Otherwise, contact the maintainer directly through the contact route listed on
+the GitHub profile or repository.
+
+## Scope
+
+Examples of in-scope issues:
+
+- command injection or unsafe shell quoting
+- install/uninstall writing outside the requested target repo
+- scripts deleting, rewriting, pushing, or overwriting work without approval
+- private paths, transcripts, or secrets being copied into a public install
+- incorrect conflict detection during bootstrap or install
+
+Examples that are usually out of scope:
+
+- a target repo's own product code
+- local machine compromise unrelated to this harness
+- unsupported modifications made after installation
+
+## Safety Principle
+
+The harness should prefer refusal over surprise. If a script cannot prove that
+an operation is scoped, reversible, or explicitly approved, it should stop and
+explain the boundary rather than guessing.