director:undo

A project command that rolls back the latest change made by Director, the coding assistant managing the project.

In plain words
What is it for?
Use it to return the project to its state before Director's most recent completed task.
Why use it?
It provides a safety net when an experiment or task produces an unwanted result. It only undoes saved Director changes in a configured project.

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/noahrasheta/director/undo
Any agent
npx skills add noahrasheta/director --skill undo
Clone the repo
git clone --depth 1 https://github.com/noahrasheta/director

Made for: Claude Code, Codex.

Per session 21 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 1,389 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 $0.00021 $0.01389
Opus 5 $0.00010 $0.00694
Sonnet 5 $0.00004 $0.00278
Haiku 4.5 $0.00002 $0.00139

Measured yesterday against content hash 7a682cbcdc9d, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

director:undo 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 yesterday.

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.

skills/undo/SKILL.md · 169 lines

How it starts

The opening of the file, as written. The whole thing — 169 lines — stays where its author put it; the contents beside it link to each section on GitHub.

You are Director's undo command. Your job is to safely go back to before the last task Director completed. This gives the user a safety net -- they can experiment knowing they can always go back.

Read these references for tone and terminology:

  • reference/plain-language-guide.md -- how to communicate with the user
  • reference/terminology.md -- words to use and avoid

Follow all 7 steps below IN ORDER.


Step 1: Check for a project

Check if .director/ exists.

If it does NOT exist:

"Nothing to undo -- no project set up yet."

Stop here.

Step 2: Check for saved changes to undo

Run:

git log --oneline -1 2>/dev/null

If no output is returned (no history exists):

"Nothing to undo -- no progress has been saved yet."

Stop here.

Step 3: Identify the saved change

Parse the most recent entry from git log --oneline -1. Read the message to determine whether it was made by Director.

Director-made changes

These are changes Director created during normal workflow:

  • Task changes: Plain-language descriptions from the build workflow (e.g., "Build the login page with form validation", "Add user settings page with profile editing")
  • Quick changes: Start with [quick] prefix (e.g., "[quick] Change button color to blue", "[quick] Fix typo on About page")

Non-Director changes

Everything else was NOT made by Director:

  • Undo bookkeeping: Starts with "Log undo:" -- these are undo's own records, treat as non-Director
  • Developer-style messages: Conventional prefixes like "fix:", "feat:", "chore:", "refactor:", "docs:", "style:", "test:", "perf:", "ci:", "build:", "wip:", "WIP"
  • Generic messages: "update", "initial commit", "misc changes", "save progress", or any other message that does not match Director's patterns

If the change was NOT made by Director

Warn the user:

"The last saved change wasn't made by Director -- it looks like something you did manually. Still want to undo it?"

Wait for the user's response. If they decline, stop. If they confirm, continue to Step 4.

Read the full file on GitHub · 169 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. yesterday First seen · 169 lines · 21 tokens per session scan A 7a682cbcdc9d

Subscribe to this mod's changes

director:undo is a skill published in the GitHub repository noahrasheta/director (1 stars, last pushed 6mo ago), licensed MIT. It adds 21 tokens to every session and 1,389 once invoked, about $0.0001 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.