MCPInstall

pulumi-upgrade-provider

by pulumi OfficialAI & Agents

Automate Pulumi provider repo upgrades with the upgrade-provider tool. Use when upgrading a pulumi provider repository to a new upstream version, running upgrade-provider, and addressing its common failure modes like patch conflicts or missing module mappings.

Installs
694
Last 24h
—
Trending rank
—
Security
Not audited

Install

Works with every agent the skills CLI supports.

npx skills add https://github.com/pulumi/agent-skills --skill pulumi-upgrade-provider
  1. 01

    Install

    Run the command in your project. The CLI asks which agents to install for.

  2. 02

    Describe the task

    Ask your agent for something this skill covers. It loads pulumi-upgrade-provider automatically when relevant.

  3. 03

    Review the result

    The agent follows the SKILL.md steps, scripts and references bundled with the skill.

SKILL.md

The instructions your agent reads when the skill is triggered.

Pulumi Upgrade Provider

Overview

Run upgrade-provider --no-submit, fix known failures, and rerun until success. During the run loop, the tool owns branch and generated-commit state but does not push or mutate GitHub. After success, review and correct the local result before submitting it using the plan printed by the tool.

Preflight

  1. Record existing changes as the pre-upgrade baseline:
baseline_file="${TMPDIR:-/tmp}/upgrade-provider-$(basename "$PWD")-baseline.txt"
git status --short --untracked-files=all | tee "$baseline_file"

Existing changes do not block the upgrade, but the tool stages broadly and may include them in generated commits. If practical, isolate clearly unrelated work in another branch or worktree, move local-only files, or add untracked local artifacts to .git/info/exclude. Otherwise proceed and carry this baseline into the final review. Do not add a repository .gitignore rule unless the path should be ignored for all contributors.

  1. Keep the run log outside the repository so the tool cannot stage it.

Run Loop

  1. Run from repo root:
log_file="${TMPDIR:-/tmp}/upgrade-provider-$(basename "$PWD").log"
upgrade-provider $ORG/$REPO --repo-path . --no-submit > "$log_file" 2> /dev/null

Preserve --no-submit on every retry.

  1. Wait for completion (can take up to 10 minutes).
  2. Check for errors by scanning the log for lines starting with error: :
if command -v rg >/dev/null 2>&1; then
  rg -n '^error: ' "$log_file" || true
else
  grep -n '^error: ' "$log_file" || true
fi
  1. On a successful exit, first check whether the tool found that no upgrade was required:
if grep -Fq 'No actions needed' "$log_file"; then
  echo 'No actions needed'
fi

If present, report that the provider is already current or the requested target is not an upgrade, then stop successfully. There is no local upgrade to review or submit.

  1. Otherwise, extract and read the complete local-completion report from the captured stdout:
awk '/^Upgrade completed locally;/{found=1} found' "$log_file"

This includes the proposed PR body and metadata, review commands, and every skipped submission action.

  1. If failed, fix using this skill's references/upgrade-provider-errors.md (from the skill folder, not the repo), then rerun with --no-submit. If upstream migrated from Terraform Plugin SDKv2 to Plugin Framework, use the migration-guide routing in that reference instead of treating the framework change alone as an architectural blocker. For upstream go get failures involving ignored replace directives or unknown revision v0.0.0, rerun with --target-version after applying the documented provider/go.mod replacements; preserve the original major/non-major intent and add --major only for actual major version upgrades.
  2. If a fix requires creating/amending/removing/rebasing patches, use the upstream-patches skill for the patch workflow. Interrupted patch workflows must be checked in before rerunning.
  3. If you fixed a conflict, report exact edits (file paths + concrete changes or preserved intent).
  4. Repeat until the tool prints No actions needed or the local-completion report and proposed PR plan. This run must not push or mutate GitHub; a branch or PR from an earlier run may already exist.

When to Stop and Report Failure

Stop iterating and report failure if any of these conditions are met:

  1. Command not found (exit code 127): The upgrade-provider tool is not in PATH.
  2. Same error 3 times: You've attempted to fix the same error 3 times without success.
  3. Unknown error pattern: The error is not covered in references/upgrade-provider-errors.md and you cannot determine a safe fix.
  4. Ambiguous patch state: You cannot prove that every patch was applied and the target rebase completed.
  5. Requires human judgment: The fix needs user input, such as choosing between multiple valid implementation approaches or making an architectural decision about module organization. Upstream breaking changes, deprecations, or framework migrations do not by themselves require stopping; stop only when implementing the upgrade presents an unresolved choice.

When stopping, report:

  1. The error(s) encountered.
  2. What fixes were attempted (with file paths and changes).
  3. Why human intervention is needed.
  4. Any partial progress.

Review the Local Upgrade

Use the tool's final report as the source of truth for the base branch, working branch, proposed PR title/body/metadata, and skipped submission actions.

  1. Review a bounded summary first:
default_branch=$(gh repo view --json defaultBranchRef --jq .defaultBranchRef.name)
base_ref="origin/$default_branch"
git status --short --untracked-files=all
git log --oneline "$base_ref"..HEAD
git diff --shortstat "$base_ref"...HEAD
git diff --name-only "$base_ref"...HEAD | awk -F/ '{print $1}' | sort | uniq -c

Provider upgrades can change thousands of generated SDK files. Do not print the full branch diff by default. Use path-limited git diff "$base_ref"...HEAD -- <paths> commands to inspect source, configuration, patch, schema, and unexpected directories. For generated SDKs, inspect summaries or targeted files unless a full diff is necessary.

Cross-check the final branch against the pre-upgrade baseline:

baseline_file="${TMPDIR:-/tmp}/upgrade-provider-$(basename "$PWD")-baseline.txt"
[ ! -f "$baseline_file" ] || cat "$baseline_file"

For every pre-existing path included in the branch, decide whether it belongs in the upgrade PR. Keep intentional paths; move unrelated work to another branch/worktree or remove it from the unpublished upgrade history before pushing. A separate commit on the same branch isolates history but does not remove the path from the PR, and unstaging alone is insufficient after the tool has committed it. Revoke any exposed credential even if the history is rewritten.

  1. MUST audit generated doc replacements for unresolved placeholders before publication:
if [ -f provider/replacements.json ]; then
  if command -v rg >/dev/null 2>&1; then
    rg -n '"new":.*TODO|TODO' provider/replacements.json || true
  else
    grep -En '"new":.*TODO|TODO' provider/replacements.json || true
  fi
fi

If any TODO is found in provider/replacements.json:

  • Treat it as a pre-submission blocker; replacement values render into generated docs.
  • Inspect each old/new pair and replace TODO with concrete Pulumi-facing wording, usually Pulumi, this provider, or the provider.
  • Run focused validation if the repo has the test:
cd provider && go test -v -run TestReplacementDoesNotIncludeTodos .
  1. If the upgrade changed patches, review the applied commits and then exit checkout mode:
./scripts/upstream.sh checkout
git -C upstream log --oneline pulumi/checkout-base..pulumi/patch-checkout
# Inspect relevant commits with git -C upstream show <sha>.
./scripts/upstream.sh check_in

List commit SHAs/titles, summarize each intent, and call out behavioral changes or risks. check_in is mandatory after review; leaving upstream on pulumi/patch-checkout blocks the next safe tool run.

  1. Run focused validation for every manual fix and add focused follow-up commits.
  2. Require a clean working tree before submission.

Submit the Reviewed Upgrade

The local-completion report lists the proposed PR metadata and every skipped action. MUST reproduce all applicable actions, not only PR creation:

  1. Preserve the proposed title, body, release label, reviewers, and assignee. Never decide, infer, or change the release label: provider upgrades follow the upstream version transition, so the label represents that version delta rather than the agent's assessment of whether individual changes are breaking. For example, upstream v0.9.0 to v0.10.0 remains a minor upgrade even when it contains breaking behavioral changes. Use exactly the label emitted by upgrade-provider, report any concern to reviewers without changing the metadata, and never add --major unless the upstream target crosses the current upstream major version.
  2. If fixes unblocked the upgrade, append this section to the proposed body before submission when practical:
---

### Fixes applied to unblock upgrade

- <list concrete unblocker edits here, with file paths and intent>
  1. Push only after review. The tool reports an unconditional force-push; manual submission intentionally uses the safer lease-protected equivalent unless the user explicitly approves overwriting newer remote work:
branch=$(git branch --show-current)
git push --set-upstream origin "$branch" --force-with-lease
  1. Use gh to create or update the PR as described by the report.
  2. Assign listed upgrade issues to the PR assignee when requested.
  3. For bridge upgrades, close listed superseded bridge PRs and comment with the replacement PR URL.
  4. Fetch and report the final PR URL:
gh pr view --json url --jq .url || gh pr list --head "$(git branch --show-current)" --json url --jq '.[0].url'

If the fixes section was not included before submission, append it without replacing the existing body:

repo=$(gh repo view --json nameWithOwner --jq .nameWithOwner)
pr_number=$(gh pr view --json number --jq .number)
gh pr view --json body --jq .body > /tmp/pr_body.txt

cat <<'EOF' >> /tmp/pr_body.txt

---

### Fixes applied to unblock upgrade

- <list concrete unblocker edits here, with file paths and intent>
EOF

gh api -X PATCH "repos/$repo/pulls/$pr_number" --raw-field body="$(cat /tmp/pr_body.txt)"

Use REST (gh api) instead of gh pr edit to avoid GraphQL project-card errors. Keep existing body content; only append.

Notes

  • git rebase --continue --no-edit is not supported in older git versions. Use git rebase --continue and accept the existing commit message.
  • To avoid the editor prompt during git rebase --continue, run it with GIT_EDITOR=true (or GIT_EDITOR=:).

Guardrails

  • Never commit, push, or create branches manually during the upgrade-provider run loop.
  • After --no-submit succeeds, focused correction commits and manual submission are permitted.
  • Never use ./scripts/upstream.sh init -f unless intentionally discarding interrupted patch work; it is destructive.
  • Do not leave upstream in checkout or an active Git operation before submission.
  • Do not stash changes; use a clean worktree or remove unrelated artifacts instead.

References

  • Use this skill's references/upgrade-provider-errors.md (from the skill folder, not the repo) for interrupted patch state, patch conflict, SDKv2-to-Plugin-Framework migration, ignored upstream replacement, vendored upstream dependency, .NET duplicate file, and new module mapping fixes.

FAQ

What does the pulumi-upgrade-provider skill do?

Automate Pulumi provider repo upgrades with the upgrade-provider tool. Use when upgrading a pulumi provider repository to a new upstream version, running upgrade-provider, and addressing its common failure modes like patch conflicts or missing module mappings.

How do I install pulumi-upgrade-provider?

Run npx skills add https://github.com/pulumi/agent-skills --skill pulumi-upgrade-provider in your project. The skills CLI detects the agents you use and copies the skill into the right folder. For Claude Code you can also place the skill folder at ~/.claude/skills/pulumi-upgrade-provider/ (personal) or .claude/skills/pulumi-upgrade-provider/ (project).

Which AI agents can use pulumi-upgrade-provider?

Any agent that supports the open Agent Skills format (a folder with a SKILL.md file), including Claude Code, Codex, Cursor, GitHub Copilot, Gemini CLI, Windsurf, Cline, OpenCode and more.

Is pulumi-upgrade-provider safe to install?

No partner security audit is available yet. Skills can run commands and scripts, so read the SKILL.md and any bundled files before installing.

Who maintains pulumi-upgrade-provider?

It is published by pulumi in pulumi/agent-skills, and it is an official skill from the maker of the technology it covers. It has 694 recorded installs.

More from pulumi/agent-skills

View all →

pulumi-best-practices

pulumi/agent-skills

Load when the user is writing, reviewing, or debugging Pulumi TypeScript/Python programs; asks about Output or apply() usage; wants to create ComponentResource classes; needs to refactor resources without destroying them (aliases); is setting up secrets or config; or is configuring a pulumi preview/up CI workflow. Also load for questions about resource dependency order, parent/child resource relationships, or pulumi.interpolate.

OfficialCode Quality & Workflow2.5K installs

pulumi-arm-to-pulumi

pulumi/agent-skills

Convert or migrate Azure ARM (Azure Resource Manager) templates, Bicep templates, or code to Pulumi, including importing existing Azure resources. This skill MUST be loaded whenever a user requests migration, conversion, or import of ARM templates, Bicep templates, ARM code, Bicep code, or Azure resources to Pulumi.

OfficialDevOps & Cloud2.1K installs

pulumi-component

pulumi/agent-skills

Guide for authoring Pulumi ComponentResource classes. Use when creating reusable infrastructure components, designing component interfaces, setting up multi-language support, or distributing component packages.

OfficialFrontend & Design1.5K installs

pulumi-esc

pulumi/agent-skills

Guidance for working with Pulumi ESC (Environments, Secrets, and Configuration). Use when users ask about managing secrets, configuration, environments, short-term credentials, configuring OIDC for AWS, Azure, GCP, integrating with secret stores (AWS Secrets Manager, Azure Key Vault, HashiCorp Vault, 1Password), or using ESC with Pulumi stacks.

OfficialDevOps & Cloud1.4K installs

pulumi-automation-api

pulumi/agent-skills

Load this skill when a user asks how to run Pulumi programmatically, embed Pulumi in an application, orchestrate multiple stacks in code, build a self-service infrastructure portal, replace pulumi CLI shell scripts with code, or use the Pulumi Automation API (LocalWorkspace, createOrSelectStack, inline programs). Also load for questions about multi-stack sequencing, parallel deployments, or passing outputs between stacks via code.

OfficialBackend & APIs1.2K installs

pulumi-terraform-to-pulumi

pulumi/agent-skills

Migrate Terraform/OpenTofu projects to Pulumi, including translating HCL source code and/or importing Terraform state into a Pulumi stack. Use when a user wants to convert Terraform to Pulumi, migrate from HCL, or import tfstate into Pulumi. Do NOT trigger for general Terraform-vs-Pulumi comparisons or questions about using both tools side-by-side.

OfficialDevOps & Cloud961 installs

Popular ai & agents skills

View category →

google-agents-cli-publish

google/agents-cli

This skill should be used when the user wants to "publish an agent", "publish my ADK agent", "register an agent with Gemini", "publish to Gemini" (formerly Gemini Enterprise), or needs guidance on the agents-cli publish gemini command. Also use when the user wants to "manage agents in Agent Registry", "list/update/delete registered agents", or "register an MCP server". Covers ADK vs A2A registration modes, programmatic and interactive usage, flag reference, auto-detection from deployment metadata, Agent Registry fleet management, and troubleshooting. Part of the agents-cli skills suite. Do NOT use for deployment (use google-agents-cli-deploy).

AI & Agents513.9K installs

google-agents-cli-eval

google/agents-cli

This skill should be used when the user wants to "run an evaluation", "evaluate my agent", "evaluate my ADK agent", "write an eval dataset", "analyze eval failures", "compare eval results", "optimize agent", or needs guidance on the Gemini platform eval methodology and the Quality Flywheel. Covers eval metrics, dataset schema, LLM-as-judge scoring, and common failure causes. Applies to any agents-cli project, whatever framework the agent is written in. Do NOT use for agent API code patterns (ADK: use google-agents-cli-adk-code), deployment (use google-agents-cli-deploy), or project scaffolding (use google-agents-cli-scaffold).

AI & Agents513.5K installs

skill-creator

anthropics/skills

Create new skills, modify and improve existing skills, and measure skill performance. Use when users want to create a skill from scratch, edit, or optimize an existing skill, run evals to test a skill, benchmark skill performance with variance analysis, or optimize a skill's description for better triggering accuracy.

OfficialAI & Agents416.4K installs

codex-pet

prime-skills/runcomfy-agent-skills

Codex Pet generator on RunComfy. Build a Codex-compatible Codex Pet spritesheet.webp + pet.json from a single reference image, drop it into ${CODEXHOME:-$HOME/.codex}/pets/ / and Codex picks it up as a custom Codex Pet next to the 8 built-ins. This skill produces the exact Codex Pet atlas Codex expects (1536x1872 PNG/WebP, 8 cols x 9 rows, 192x208 cells, 9 animation states — idle, running-right, running-left, waving, jumping, failed, waiting, running, review). Calls OpenAI GPT Image 2 edit ONCE via the local RunComfy CLI as runcomfy run openai/gpt-image-2/edit to produce a canonical Codex Pet pose, then assembles all 9 animation rows programmatically with ImageMagick micro-transforms — no Codex Pro, no $imagegen, no OPENAIAPIKEY required, only RUNCOMFYTOKEN. Triggers on "codex pet", "create codex pet", "make codex pet", "hatch codex pet", "/hatch image", "desktop pet codex", "codex pets", "spritesheet.webp", or any explicit ask to build a custom pet for OpenAI Codex.

AI & Agents402.1K installs

full-output-enforcement

leonxlnx/taste-skill

Overrides default LLM truncation behavior. Enforces complete code generation, bans placeholder patterns, and handles token-limit splits cleanly. Apply to any task requiring exhaustive, unabridged output.

AI & Agents401K installs

executing-plans

obra/superpowers

Use when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available

AI & Agents245.7K installs