Agent skill

Rollback

by aiblueprinthq in aiblueprinthq/ai-blueprint

Plan a safe history-preserving reversal of a completed feature from its archive and exact commit, including later-dependency risk, then stop before implementation.

MITAuto-check passedDevOps & Cloud

Install Rollback

skills CLI
$ npx skills add aiblueprinthq/ai-blueprint --skill rollback -a claude-code

Project install by default; add -g for ~/.claude/skills/.

GitHub CLI
$ gh skill install aiblueprinthq/ai-blueprint rollback --agent claude-code

Project scope by default; add --scope user for a personal install. Needs GitHub CLI 2.90.0 or later (public preview).

Manual copy
$ git clone --depth 1 https://github.com/aiblueprinthq/ai-blueprint.git skills-src && mkdir -p .claude/skills && cp -r skills-src/.agents/skills/rollback .claude/skills/rollback && rm -rf skills-src

Use ~/.claude/skills/ instead of .claude/skills for a personal install. The folder must contain SKILL.md.

Claude Code skills documentation · loads skills from .claude/skills/

Facts

Skill name
rollback
GitHub stars
463
Token cost
~2.3k tokens
SKILL.md length
1,237 words
Files
3
Skills in repo
20
Repo updated
First seen
Licence
MIT

At a glance

Plan a safe history-preserving reversal of a completed feature from its archive and exact commit, including later-dependency risk, then stop before implementation.

  • Works in 5 steps: preflight → resolve the exact feature → separate product changes from Blueprint… → …
  • Requests to undo
  • SKILL.md covers Input, Step 0 - preflight, Step 1 - resolve the exact… and Step 2 - separate product…, plus 4 more sections
  • Calls git

What it does

Rollback is an agent skill from aiblueprinthq/ai-blueprint. Plan a safe history-preserving reversal of a completed feature from its archive and exact commit, including later-dependency risk, then stop before implementation. Use for /rollback or requests to undo, remove, or reverse a completed feature.

Its SKILL.md is about 2.3k tokens, which your agent loads only when the skill is triggered. The skill folder holds 4 other files (for example `agents/openai.yaml` and `reference/rollback-spec-template.md`).

It sits in DevOps & Cloud. It works with Git. The repository describes itself as: A file-backed, spec-driven AI coding workflow framework for building real software while staying in control. The licence is MIT.

When your agent uses it

  • Requests to undo
  • Reverse a completed feature

Example prompts

  • “/rollback”

Workflow steps

5 steps, taken from the step headings in SKILL.md.

  1. preflight
  2. resolve the exact feature
  3. separate product changes from Blueprint history
  4. review later-change risk
  5. write the rollback spec

What it can do on your machine

Read from SKILL.md and the folder at commit 96222b7. It shows what the files ask for, not the result of running them.

  • Tool permissions

    Pre-approves nothing: there is no allowed-tools line, so your agent's usual permission prompts apply.

    From allowed-tools in the SKILL.md frontmatter.

  • Runs code

    Shell commands in SKILL.md call:

    • git

    From the folder's file list and the shell code blocks in SKILL.md.

  • Network

    No URLs in SKILL.md. Its commands use git, which can reach the network depending on how they are called.

    From URLs in SKILL.md, links to its own repository left out.

  • Credentials

    Names no API keys, tokens, secrets or passwords.

    From names ending in _API_KEY, _TOKEN, _SECRET, _KEY or _PASSWORD in SKILL.md.

Context cost

Rollback loads about 2.3k tokens when it runs. Until then it costs about 63 tokens; SKILL.md has 1,237 words of instructions outside code blocks.

Always · name and description, kept in context so the agent knows when to use it
~63
When it runs · the whole SKILL.md, loaded when a task matches
~2.3k

Estimates: characters ÷ 4, the usual rule of thumb; real counts depend on the model's tokenizer. Scripts and assets cost tokens only if the agent reads them.

Safety

Auto-check passed

The automated check found no risky patterns in SKILL.md.

Automated static check — not a guarantee. Review scripts before installing. It scans the text of SKILL.md for risky patterns (piping downloads into a shell, reading credential files, hidden Unicode, destructive commands); files beside SKILL.md are not scanned.

SKILL.md

The full file from aiblueprinthq/ai-blueprint at commit 96222b7, republished under its MIT licence (© aiblueprinthq). 1,237 words, ~2,281 tokens.

Download SKILL.mdSave it as .claude/skills/rollback/SKILL.md (or your agent's skills folder). This skill also uses 2 other files; get the full folder from GitHub.
name
rollback
description
Plan a safe history-preserving reversal of a completed feature from its archive and exact commit, including later-dependency risk, then stop before implementation. Use for /rollback or requests to undo, remove, or reverse a completed feature.

rollback - safely reverse a completed feature

Context reuse: Reuse any required file already loaded in project instructions or the current session. Read it again only if absent, changed, or exact current bytes or line references are needed.

First action: Before project inspection, preflight, or any other tool call, publish running to blueprint/.state/run.json using the dashboard activity contract in AGENTS.md.

Where this sits in the workflow:

completed feature + git history  ->  [rollback]  ->  /implement  ->  /check  ->  /complete
(archive + squashed commit)           (risk review     (reverse       (prove)     (log + merge)
                                       + spec)          product diff)

This skill plans a rollback. It does not change product code, create a branch, commit, merge, or push. It identifies the completed feature and its exact git commit, checks what changed afterward, writes a guarded rollback spec, then stops for review. /implement performs the reversal only after the user approves that spec.

Input

A completed feature by build-plan number, name, or archive path, plus an optional reason. Examples:

/rollback 4 because the new export flow is corrupting files
/rollback "PDF export"
/rollback blueprint/history/features/04-pdf-export.md

With no target, list a short set of recent completed feature archives and ask the user to choose. Never silently pick the latest feature. If the reason is missing, ask for one before writing the spec because the rollback archive must explain why the feature was removed.

Step 0 - preflight

Read AGENTS.md, blueprint/build-plan.md, blueprint/context/current-feature.md, the completed feature archives, and git state.

Stop before writing when:

  • the directory is not a git repository
  • current-feature.md already holds active work. Never replace another worker's active spec; parallel work uses one clone or Git worktree per work item, with a separate branch in each checkout
  • the working tree is dirty, including unrelated untracked work
  • the current branch is not the local main or default branch
  • the target is not a checked build-plan feature with a matching archive
  • the archive or its introducing commit cannot be identified unambiguously

Do not stash, discard, fetch, pull, switch branches, or clean the tree from this skill.

Step 1 - resolve the exact feature

Resolve the requested number or name to one checked build-plan ID; for an archive path, read its exact spec identity first and match that checked item. Then gather all exact-ID feature archive candidates using ../feature/reference/build-history.md. Normalize the stable ID, retaining its letter, and read attempts from verified spec metadata/history proof, never arbitrary title or filename suffixes.

For each candidate, require its unique introducing commit on the local default and prove the archive blob and mode have not changed since that addition. Exclude builds with a proven completed reachable rollback naming that exact archive and introducing commit. The checked item must have exactly one unreversed build. An explicit archive request stays bound to that exact path: reject it if already reversed, rather than silently choosing another build. Ambiguous, orphan, overwritten, multiply-added, or missing records stop with their exact paths and missing proof. Do not select by newest commit, filename order, or rollback date.

Confirm the selected commit's diff is consistent with the requested feature. If the target is a merge commit, stop before Step 2 and before writing or changing blueprint/context/current-feature.md. Do not record a target parent or choose a mainline. Publish blocked to blueprint/.state/run.json, explain that Blueprint cannot safely infer which merge parent represents the pre-feature state, and include the exact /rollback command the user can rerun after choosing a safe remediation or mainline strategy. /implement retains its merge-target stop as defense in depth. If the archive was never committed, explain that git cannot reconstruct a safe rollback from it.

Step 2 - separate product changes from Blueprint history

Inspect the target commit and build the product-path set from the files it changed. Exclude these protected workflow paths:

  • .agents/**
  • .claude/**
  • blueprint/**
  • AGENTS.md
  • CLAUDE.md
  • prototypes/**

The rollback must preserve the original feature archive, later planning changes, the active rollback spec, adapter skills, and throwaway prototype history. Root README.md and ordinary app docs are product files unless the project says otherwise.

If no product paths remain, stop. Do not create an empty rollback that only rewrites Blueprint records.

Step 3 - review later-change risk

Inspect every commit after the target through HEAD that touches one of the product paths. Also read later completed feature specs when they mention the target, its contracts, routes, data, or files.

Classify the result:

  • No overlap - no later commit touched the target product paths.
  • Overlap, likely compatible - later edits touched the same paths, but the target change can be reversed without removing their behavior.
  • Dependency risk - later work appears to require the target's API, schema, route, component, or data.
  • Blocked - safe behavior after reversal is unclear, data migration would be destructive, or a cascading rollback would be required.

Path overlap is a warning signal, not proof of dependency. Explain the concrete later commit and contract involved. Never silently cascade into reverting other features. For a blocked case, stop and ask the user to choose a narrower remediation or explicitly plan the dependent rollbacks.

Show full SKILL.md (421 more words)Show less

Step 4 - write the rollback spec

Write blueprint/context/current-feature.md using reference/rollback-spec-template.md. Its first heading must be exactly # Rollback: Feature <id> - <title>, and it must retain the existing **Type:** Rollback contract. Fill in:

  • the full rollback branch from the configured prefix plus the rollback title in lowercase kebab-case, appending --build-N for target attempt N > 1 proved above
  • target feature and archive
  • target commit and parent commit as full 40-character SHA values
  • user's reason
  • product paths introduced or changed by the target
  • protected workflow paths
  • later commits reviewed and the risk classification
  • compatibility work that is allowed, if any
  • exact verification commands and observable removal criteria

Before freezing a new rollback spec, apply the archive-path and branch-availability checks in ../feature/reference/build-history.md to its planned branch and YYYY-MM-DD-<exact-target-archive-stem>.md destination. Retain --build-N in the target stem when present. A collision stops for an explicit decision; never renumber the target or silently rename reviewed work. Existing completion recovery runs first.

The first build step must apply the target commit's product diff in reverse using the guarded Type: Rollback behavior in /implement. Later steps may repair only the specific downstream compatibility issues named in the spec. Do not use a rollback as permission for unrelated cleanup.

Older installations may not have blueprint/history/rollbacks/ yet because updates preserve user history. That is not a planning blocker; /complete creates the directory when it archives the approved rollback.

Red-team the draft before presenting it:

  • does it preserve all Blueprint history and later plan changes?
  • could it remove data or require a destructive migration?
  • does later code import or call something the target introduced?
  • are the removal criteria observable rather than phrased as "feature gone"?
  • can each compatibility edit be reviewed separately?

Tighten the spec, then stop. Summarize the target commit, affected product paths, later-change risk, and what the critique changed. Tell the user to review the spec, then run /implement to create the rollback branch and apply it.

Rules

  • Preserve history. Never delete or rewrite the original feature archive.
  • Plan only. This skill writes the rollback spec and nothing else.
  • One completed feature per rollback.
  • Record both the target commit and its parent as full 40-character SHA values. Each value must resolve to the recorded commit in the current repository.
  • Never use git reset --hard, force-push, history rewriting, or broad file restoration.
  • Never infer permission to cascade into later features or destroy stored data.
  • A rollback still uses /implement, /check, and /complete review gates.

Formatting

Format the output to match blueprint/context/ai-interaction.md: concise, scannable markdown with a small risk table when later commits overlap.

© aiblueprinthq, MIT. Rendered from Markdown: HTML in the file is shown as text, images as links, and headings moved down two levels. Raw file

Files

SKILL.md and 2 other files in .agents/skills/rollback of aiblueprinthq/ai-blueprint.

  • SKILL.md
  • agents/openai.yaml
  • reference/rollback-spec-template.md

Open the folder on GitHubat commit 96222b7

Compare with similar skills

Rollback next to the 5 skills that share the most tags, products or categories with it. Stars are the repository's; “used in” counts other GitHub owners with a copy.

Rollback compared with similar skills
SkillStarsUsed inTokensAuto-checkLicenceRepo updated
Rollback this skillaiblueprinthq/ai-blueprint463—~2.3kAutomated safety check: PassMIT
Repo Mirror Sourcesnetdata/netdata81k—~1.2kAutomated safety check: NotesGPL-3.0
CI Failure Triage and RepairChachamaru127/claude-code-harness3.2k1 repos~1.1kAutomated safety check: NotesMIT
Bfe Rd Workflowbfenetworks/bfe6.3k—~1.3kAutomated safety check: PassApache-2.0
Ssh Skillbadseal/ssh-skill537—~2.4kAutomated safety check: NotesNone
GreptimeDB Release RunbookGreptimeTeam/greptimedb6.7k—~1.4kAutomated safety check: PassApache-2.0

Similar skills

  • Repo Mirror Sources

    netdata/netdata

    Inspect Netdata-org source checkouts under NETDATAREPOSDIR, or set up and synchronize that mirror when requested.

    81k GitHub stars~1.2k tokensUpdated today
    DevOps & CloudAuto-check: notes
  • CI Failure Triage and Repair

    Chachamaru127/claude-code-harness

    Diagnoses failing CI pipelines and tests, deciding first whether the test or the implementation is at fault, and hands hard cases to a dedicated fixer subagent.

    3.2k GitHub starsUsed in 1 repo~1.1k tokens
    DevOps & CloudAuto-check: notes
  • Bfe Rd Workflow

    bfenetworks/bfe

    引导用户在 bfe 代码库中完成一次完整的功能研发流程,包括需求对齐、文档修改、代码实现、集成测试与回归验证. An agent skill from bfenetworks/bfe.

    6.3k GitHub stars~1.3k tokensUpdated today
    DevOps & CloudAuto-check passed
  • Ssh Skill

    badseal/ssh-skill

    A skill your agent uses when a task requires SSH or SCP/SFTP behavior, a remote server, server alias/IP/hostname/user@host, bastion or jump-host access, remote command execution, upload/download…

    537 GitHub stars~2.4k tokensUpdated 1 mo ago
    DevOps & CloudAuto-check: notes
  • GreptimeDB Release Runbook

    GreptimeTeam/greptimedb

    Runbook for publishing a GreptimeDB version: pick the release branch, verify the Cargo version, then tag, create the GitHub release and open the docs note PR.

    6.7k GitHub stars~1.4k tokensUpdated today
    DevOps & CloudAuto-check passed
  • Crabbox Quickstart

    openclaw/crabbox

    Gets you running your repository's tests in a disposable Docker or Podman container on your own machine with Crabbox, with no account and no cloud spend.

    1.5k GitHub stars~1.6k tokensUpdated today
    DevOps & CloudAuto-check: notes

More from aiblueprinthq/ai-blueprint

All 20 skills in this repo
  • Adopt

    aiblueprinthq/ai-blueprint

    Adopt Blueprint into an existing brownfield codebase by surveying shipped behavior and generating plans, standards, commands, adapter choices, and visibility setup.

    463 GitHub stars~2.7k tokensUpdated today
    Auto-check passed
  • CI

    aiblueprinthq/ai-blueprint

    Set up or normalize one project Verify command and matching GitHub Actions checks while preserving existing CI, with an optional local pre-push hook.

    463 GitHub stars~2.2k tokensUpdated today
    Auto-check passed
  • Doctor

    aiblueprinthq/ai-blueprint

    Run a Blueprint health and context check covering setup, adapters, commands, visibility, plans, overview freshness, configuration, dashboard state, and workflow drift.

    463 GitHub stars~4.5k tokensUpdated today
    Auto-check: notes
  • Feature

    aiblueprinthq/ai-blueprint

    Turn the next, named, or numbered build-plan feature into a buildable current-feature.md spec with small steps and done-when criteria.

    463 GitHub stars~2.8k tokensUpdated today
    Auto-check passed
  • Onboard

    aiblueprinthq/ai-blueprint

    Onboard a fresh or early scaffold after Blueprint is overlaid by tuning commands, standards, adapters, visibility, and context loading.

    463 GitHub stars~4.5k tokensUpdated today
    Auto-check passed
  • Overview

    aiblueprinthq/ai-blueprint

    Validate and normalize project-plan.md and build-plan.md, then generate the durable project-overview.md used by agents.

    463 GitHub stars~3.8k tokensUpdated today
    Auto-check passed

Works with

Categories

Questions about Rollback

What does Rollback do?

Plan a safe history-preserving reversal of a completed feature from its archive and exact commit, including later-dependency risk, then stop before implementation. Rollback is an agent skill from aiblueprinthq/ai-blueprint. Plan a safe history-preserving reversal of a completed feature from its archive and exact commit, including later-dependency risk, then stop before implementation.

When should I use Rollback?

Rollback fits situations like: requests to undo; reverse a completed feature.

How do I install Rollback in Claude Code?

Run `npx skills add aiblueprinthq/ai-blueprint --skill rollback -a claude-code`. Or copy the skill folder (.agents/skills/rollback in aiblueprinthq/ai-blueprint) into .claude/skills/rollback in your project. Claude Code loads it when a task matches its description.

How do I install Rollback in Codex?

Run `npx skills add aiblueprinthq/ai-blueprint --skill rollback -a codex`. Or copy the skill folder (.agents/skills/rollback in aiblueprinthq/ai-blueprint) into .agents/skills/rollback in your project. Codex loads it when a task matches its description.

Can I use Rollback in Cursor, Gemini CLI or GitHub Copilot?

Cursor, Gemini CLI, GitHub Copilot and OpenCode also load SKILL.md folders. With the skills CLI, run `npx skills add aiblueprinthq/ai-blueprint --skill rollback -a cursor` (or -a gemini-cli, github-copilot or opencode for the others). To copy it by hand, put the folder in .cursor/skills/rollback, .gemini/skills/rollback, .github/skills/rollback and .opencode/skills/rollback in your project.

What does Rollback need to run?

Going by SKILL.md and its folder, Rollback needs the command-line tools its instructions call (git).

Does Rollback access the network?

SKILL.md contains no URLs. Its commands use git, which can reach the network depending on how they are called. This is read from the text; nothing was executed.

Is Rollback safe to install?

Our automated static check of SKILL.md found no risky patterns, such as piping downloads into a shell, reading credential files or hidden Unicode. It is not a guarantee. Review the folder before installing.

What licence does Rollback use?

Rollback is published under the MIT licence (the repository's licence). It allows redistribution, so the full SKILL.md is shown on this page.

How many tokens does Rollback use?

About 2.3k tokens (SKILL.md is roughly 9.1k characters). Agents keep only the skill's name and description in context until a task matches; then they load SKILL.md in full.

What are the alternatives to Rollback?

Skills that share tags, products or a category with Rollback: Repo Mirror Sources (netdata/netdata, 81k stars), CI Failure Triage and Repair (Chachamaru127/claude-code-harness, 3.2k stars), Bfe Rd Workflow (bfenetworks/bfe, 6.3k stars) and Ssh Skill (badseal/ssh-skill, 537 stars). The comparison table on this page puts their stars, adoption, token cost, safety result and licence side by side.

Who maintains Rollback?

aiblueprinthq (a GitHub organization) maintains it in aiblueprinthq/ai-blueprint, which has 463 GitHub stars. The repository holds 20 skills in this directory. The repository was last updated on October 8, 2026.

Source: aiblueprinthq/ai-blueprint on GitHub. Facts on this page come from the repository at the commit we read; the author's words are quoted as theirs.