plumb-git

A guide for using plumb’s guarded Git tool. Git is a version-control system, and the guide explains which commands are allowed and which need extra permission or confirmation.

In plain words
What is it for?
Use it for status, history, diffs, commits, branch changes, merges, resets, cleanup, rebases, and network operations such as push or pull.
Why use it?
It helps prevent unsafe Git operations and explains why a command was refused, while keeping repository access under plumb’s rules.

Skill for Claude CodeCodex

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add skills/plumbkit/plumb/plumb-git
Any agent
npx skills add plumbkit/plumb --skill plumb-git
Clone the repo
git clone --depth 1 https://github.com/plumbkit/plumb

Made for: Claude Code, Codex.

Per session 56 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 1,676 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5 $0.00056 $0.01676
Opus 5 $0.00028 $0.00838
Sonnet 5 $0.00011 $0.00335
Haiku 4.5 $0.00006 $0.00168

Measured 2d ago against content hash 6c895dce0474, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

plumb-git scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

internal/cli/skills/plumb-git/SKILL.md · 80 lines

How it starts

The opening of the file, as written. The whole thing — 80 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Plumb's git tool is a policy-gated wrapper, not a shell. The subcommand leads the argv, nothing is interpolated by a shell, and eight global flags that could re-target the repository or inject config are refused wherever they appear in args: -c, -C, --exec-path, --git-dir, --work-tree, --namespace, --upload-pack, --receive-pack. The one exception is switch, where -c / -C mean create-branch rather than anything global: they are rewritten to --create / --force-create before the denylist runs, and the response says so. Prefer this tool over shelling out — the shell path bypasses the policy and plumb's per-repository git lock.

Under a lean tool profile file_status and minimal_diff_review are not advertised; set [tools] profile = "full" in .plumb/config.toml to see them.

The four tiers

Tier Subcommands Gate
read status, log, diff, show, blame, shortlog, check-ignore always allowed
write add, commit, mv [git] allow_writes
destructive reset, clean, rebase, revert, cherry-pick [git] allow_destructive and a confirmation
network push, fetch, pull [git] allow_push and a confirmation

rm is refused at every tier: delete the file with delete_file, then stage the deletion with add.

Six subcommands are classified by their arguments, biased towards the safer-to-deny higher tier — so the same subcommand can land in different tiers on different calls:

  • checkout -b / -B (branch creation) is write; every other checkout is destructive, since it can discard the working tree or detach HEAD. Prefer switch for a safe branch change.
  • switch is write, but switch -f / --force / --discard-changes is destructive.
  • restore: with --staged (the index only) it is write; with --worktree, or with no flag at all, it is destructive.
  • branch: --list / -l / -a / -r / -v / --show-current / --contains / --merged, and a bare branch, are read; creating, -m / -M / --move / --copy are write; -d / -D / --delete is destructive.
  • tag: -l / --list / -n / --contains / --merged, and a bare tag, are read; creating is write; -d / --delete is destructive.
  • stash: list / show are read; a bare stash plus push / save / pop / apply / create / store are write; drop / clear are destructive; any other sub-subcommand is refused with the permitted list.

session_start prints the live policy — read it there rather than discovering a tier by being refused.

Read the full file on GitHub · 80 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 2d ago First seen · 80 lines · 56 tokens per session scan A 6c895dce0474

Subscribe to this mod's changes

plumb-git is a skill published in the GitHub repository plumbkit/plumb (4 stars, last pushed 2d ago), licensed MIT. It adds 56 tokens to every session and 1,676 once invoked, about $0.0003 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.

Related

Other skills, from other repositories