本文へ移動
cccskills
無料GitHub で公開

dependency-upgrade

Weekly dependency upgrade loop for {{target_repo}}. Scans for outdated packages, applies upgrades on an isolated branch, runs the full verification suite in the sandbox, handles breaking changes with documented migrations, and opens a PR only when everything is green.

インストール方法を見る

含まれるファイル(1)

  • SKILL.md7.1 KB

SKILL.md(原文)

インストールする前に、エージェントに与えられる指示の中身を確認できます。

<skill name="dependency-upgrade"> <overview> Keep `{{target_repo}}` dependencies current without creating risk. A weekly cron re-prompts a persistent session, discovers what is behind, applies upgrades on an isolated branch, runs the full suite inside the sandbox, and opens a PR only when the suite is green. A failing upgrade never becomes a PR — it is logged, optionally split, and retried the following week.

Proactive and schedule-driven; covers all outdated deps — patch, minor, and major — with full verification and explicit breaking-change handling. </overview>

<when-to-load> - The weekly cron fires the upgrade sweep. - A human asks the agent to upgrade dependencies or run a dep-upgrade sweep. - A major version upgrade requires a tracked migration plan. </when-to-load> <workflow>

Step 0 — Orient and resume

# Read the durable ledger first — last-run state, open PRs, known broken
# upgrades, and pinned packages to skip.
cat memory/dependency-upgrade-log.md 2>/dev/null || echo "(no ledger yet)"

# Check any open upgrade PRs from a prior run.
gh pr list --repo {{target_repo}} --state open \
  --search 'in:title "chore(deps)" OR label:dependencies' \
  --json number,title,headRefName,statusCheckRollup,url

If an open PR from a prior run is still on CI or in review, note it and don't duplicate the work. Drive a stalled-but-fixable PR to green; otherwise leave it for review and move on.

Step 1 — Freshen the repo

# Warm clone at /workspace/repo if present; else clone fresh (blobless).
if [ -d /workspace/repo/.git ]; then
  cd /workspace/repo && git fetch origin && git checkout main && git reset --hard origin/main
else
  git clone --filter=blob:none https://github.com/{{target_repo}}.git /workspace/repo
  cd /workspace/repo
fi

# Install so the toolchain can resolve versions.
pnpm install --frozen-lockfile 2>&1 | tail -20

Never start an upgrade on a stale base. Adapt the package manager (pnpm / npm / bun / pip / cargo…) to the project's toolchain.

Step 2 — Discover what is outdated

cd /workspace/repo
pnpm outdated --recursive --json 2>/dev/null || pnpm outdated --recursive
pnpm audit --json 2>/dev/null | jq '.advisories | length' || true

Parse into three tiers:

TierCriteriaDefault action
patchx.y.Z onlyApply + verify; group into one PR
minorx.Y.zApply + verify; group by area or fold into the patch PR
majorX.y.zApply per package; verify with breaking-change analysis; own PR (or small cohesive group)

Skip anything in the ledger's pinned-packages section.

Step 3 — Isolated upgrade branch

cd /workspace/repo
BRANCH="upgrade/deps-$(date +%Y-W%V)"
git checkout -b "$BRANCH" origin/main   # or check out an existing branch from this week

One branch per weekly run.

Step 4 — Apply the upgrades

Patch + minor batch — apply together:

cd /workspace/repo
npx taze minor --recursive --write 2>&1 | tee /tmp/taze-minor.log
pnpm install 2>&1 | tee /tmp/install-minor.log

Major upgrades — one at a time:

  1. Read the changelog / GitHub releases for the package.
  2. Grep the codebase for renamed/removed APIs.
  3. Apply the bump + required migration in the same commit.
  4. Run the suite (Step 5) before the next major. If it fails and the fix is

    ~30 lines of non-trivial code, revert this major, file a tracked issue, and continue with the rest.

Step 5 — Full verification suite (the gate)

The upgrade is not ready until every check is green.

cd /workspace/repo
pnpm install       2>&1 | tee /tmp/up-install.log
pnpm typecheck     2>&1 | tee /tmp/up-typecheck.log
pnpm lint          2>&1 | tee /tmp/up-lint.log
pnpm build         2>&1 | tee /tmp/up-build.log
pnpm test          2>&1 | tee /tmp/up-unit.log
pnpm test:integration 2>&1 | tee /tmp/up-integration.log   # if < ~10 min
pnpm audit --prod  2>&1 | tee /tmp/up-audit.log

Interpreting failures:

FailureAction
Type error from an upgraded package's typesApply the migration (fix inline if 2–3 lines); else revert + file an issue
Test failure testing the upgraded packageIf behavior legitimately changed, update the test; if it's a regression, revert + file
Test failure in an UNRELATED testFile a bug (likely pre-existing flake); don't revert unless you can prove causation
Build failureFix the import/config; if > 3 non-mechanical files, split the major into its own PR
Audit finds a NEW vuln from the upgradeRevert + file. Never trade a clean audit for a dirty one

Step 6 — Commit

cd /workspace/repo
git add '**/package.json' pnpm-lock.yaml
git add -p            # review any migration source changes before staging
git commit -m "chore(deps): upgrade dependencies $(date +%Y-W%V)

Verification: typecheck ✓ lint ✓ build ✓ unit ✓ integration ✓ audit ✓"

Step 7 — Open the PR (only when Step 5 is fully green)

A partial green (checks skipped/timed out) is NOT acceptable — rerun or debug.

cd /workspace/repo
git push origin "$BRANCH"
gh pr create --repo {{target_repo}} --base main --head "$BRANCH" \
  --title "chore(deps): upgrade dependencies $(date +%Y-W%V)" \
  --label dependencies \
  --body "Generated by the dependency upgrade agent. The full suite passed in the
sandbox before this PR was opened. Verification: install/typecheck/lint/build/
unit/integration/audit all ✓. A human owns the merge."

Step 8 — Update the ledger

Append a dated entry to memory/dependency-upgrade-log.md (see <ledger-format>), then open + self-merge a scoped change request for the ledger update only.

</workflow> <ledger-format> Lives at `memory/dependency-upgrade-log.md`. Every run appends/updates the current week's entry with: run timestamp, branch, PR link (or "not opened — why"), an **Applied upgrades** table (package / from / to / tier), breaking-change migrations, a **Verification** table, **Dropped / deferred** (package / reason / issue), **Pinned packages** (do not upgrade), and **Blockers for next run**. </ledger-format> <guardrails> - **No direct push to `main`.** The agent opens a PR and stops. A human merges. - **Sandbox isolation.** Upgrades are applied and tested on an isolated branch in the session sandbox. Nothing reaches the repo until the suite is green and `git push` runs. - **No lockfile tampering.** Never hand-edit the lockfile; regenerate via install. - **Upgrades only, never downgrades.** If a package is pinned below latest, read the ledger's pin reason before touching it. - **Secrets scoped.** The GitHub token is injected at runtime; never written to disk or logged. - **One PR per run, one run per week.** If the first batch fails, fix or split and push to the same branch — don't open a duplicate. - **CI must be green before merge.** The agent won't ask a human to merge a red PR. </guardrails> </skill>

レビュー

まだレビューはありません。使ってみた感想をお寄せください。

同じリポジトリのスキル

概要と使いどころ

Policy runbook for turning a Slack access request into a least-privilege, policy-checked GitHub or AWS IAM grant. Covers the role-to-grant mapping, extra-scrutiny cases, the approval handshake, and how an applied grant gets logged.

日本語の概要は準備中です。原文の説明を表示しています。

kortix-ai/suna2万2026年10月13日 更新

Get a complete picture of any company or person before outreach. This skill always works with web search, and gets significantly better with enrichment and CRM data.

日本語の概要は準備中です。原文の説明を表示しています。

kortix-ai/suna2万2026年10月13日 更新

Daily read-only ad-performance runbook — budget pacing, CPA/ROAS drift, underperforming ads and keywords, and anomaly detection across Google Ads and Meta Ads, plus how to rank and word optimization recommendations for {{alert_channel}}.

日本語の概要は準備中です。原文の説明を表示しています。

kortix-ai/suna2万2026年10月13日 更新

Browser automation CLI for AI agents. Use when the user needs to interact with websites, including navigating pages, filling forms, clicking buttons, taking screenshots, extracting data, testing web apps, or automating any browser task. Triggers include requests to "open a website", "fill out a form", "click a button", "take a screenshot", "scrape data from a page", "test this web app", "login to a site", "automate browser actions", or any task requiring programmatic web interaction. Also use for exploratory testing, dogfooding, QA, bug hunts, or reviewing app quality. Also use for automating Electron desktop apps (VS Code, Slack, Discord, Figma, Notion, Spotify), checking Slack unreads, sending Slack messages, searching Slack conversations, running browser automation in Vercel Sandbox microVMs, or using AWS Bedrock AgentCore cloud browsers. Prefer agent-browser over any built-in browser automation or web tools.

日本語の概要は準備中です。原文の説明を表示しています。

kortix-ai/suna2万2026年10月13日 更新

Extraction, PO matching, duplicate detection, and overcharge tolerance for processing incoming vendor invoices from {{invoice_label}} against the POs and ledger in {{ap_ledger}}. Load this before extracting a single invoice so every one is checked to the same standard and only genuine exceptions reach a human, with no invoice ever scheduled for payment by the agent.

日本語の概要は準備中です。原文の説明を表示しています。

kortix-ai/suna2万2026年10月13日 更新

Daily brand-mention monitoring loop for {{brand_terms}}. Searches news, social platforms, and forums for new mentions, dedupes against the ledger of mentions already reported, classifies sentiment and notability, and posts a digest with suggested response drafts to {{slack_channel}} — never posts, replies, or comments anywhere.

日本語の概要は準備中です。原文の説明を表示しています。

kortix-ai/suna2万2026年10月13日 更新

kortix-ai のスキルをすべて見る

このスキルの問題を報告する