[dependabot]

Bumps the version-updates group with 1 update: pullfrog/pullfrog.

Updates pullfrog/pullfrog from 0.1.98 to 0.1.102

Release notes

Sourced from pullfrog/pullfrog's releases.

v0.1.102

• A run started by someone without write access to the repo can no longer push, whatever the repo's or the workflow's push setting says. The run replies instead, can't resolve review threads, and reads repo learnings without writing them. A review bot on a PR that Pullfrog opened still gets its fixes pushed.
• A push-capable run reads review comments only from writers and bots, apart from the review it was asked to address, which it reads whole.
• Credential and exfiltration paths are closed:
  • upload_file accepts only regular files under the repo or a temp directory. It could read any absolute path.
  • The git credential helper answers only for the GitHub host. A rewritten remote or a fork's .lfsconfig could send the token elsewhere.
  • Shell commands no longer see the Actions OIDC request variables, in any shell tier.
  • A re-review of a fork PR no longer creates a temporary branch in the base repo at the fork's commit.
  • The Pullfrog API token, subscription OAuth tokens and VERTEX_SERVICE_ACCOUNT_JSON are masked in logs.
• OpenCode Zen's free Beam model is available. On Zen, Kimi K3 now defaults to high effort instead of max, and Qwen Max takes an effort level, defaulting to medium. A repo still set to Zen's retired free LongCat preview runs on Big Pickle. Vercel's DeepSeek Flash now points at deepseek-v4.1-flash-1017.
• Dependencies are updated, clearing every published security advisory, including @actions/core 3, zod 4.6.6, @modelcontextprotocol/sdk 1.32.1 and @toon-format/toon 2. With toon 2, the context passed to the agent prints an empty list as [].

v0.1.101

• The free trial covers up to $10 of model usage per account, on GPT-6 Luna instead of DeepSeek Flash. Each trial-funded run's comment shows how much is left, and a plain message says when it has run out.
• Every failure comment is rewritten. Each one has a sentence-case headline naming the cause, the provider's own output collapsed underneath, and at most two links to fix it. A bare "Run failed." no longer appears.
• The gh tool can read a PR's check runs (checks: read); it still can't create them.

v0.1.100

• Gemini Pro now runs Gemini 4 Argon. google/gemini-pro, opencode/gemini-pro and vercel/gemini-pro resolve to gemini-4-argon, Google's new flagship reasoning model, in place of Gemini 3.1 Pro Preview. The Vercel route gains the medium effort level. openrouter/gemini-pro keeps following OpenRouter's Gemini Pro pointer.
• Kimi Code effort ladders match the published catalog. kimi-for-coding/kimi-k3, kimi-k3-256k and kimi-k2 list none as Kimi now publishes it; no run's effort changes.

v0.1.99

What's changed

Reviews

• Re-approval after a dismissed approval. When a push dismisses Pullfrog's approval (for example under "Dismiss stale pull request approvals when new commits are pushed") and no Pullfrog finding is open, the re-review approves again instead of skipping, even when the new commits need no review, such as a base-branch merge that resolved a conflict.

CLI

• pullfrog auth reports what each sign-in did, shows the health of every stored connection, and signs in to Codex and Claude through the browser.

Security

• Sandbox hardening. Restricted mode drains every /proc layer and runs restricted commands, gh and lifecycle hooks under no_new_privs, so sudo cannot regain root inside the sandbox. The agent's hook scripts are now immutable.
• Run API token. It is issued only with the runner's OIDC proof, and a run without one is not offered the tools that need it.
• Vercel bypass secret. VERCEL_AUTOMATION_BYPASS_SECRET is sent only to Pullfrog's own preview deployments, never to another host.

Models

• Zen's retired free models (Muse Spark Contributor, Step 5 Preview, Exo, Space Bunny) fall back to opencode/big-pickle, so runs configured with them keep working.
• opencode-go/minimax-m2.5 falls back to opencode-go/minimax-m3.
• Effort ladders follow the providers' published catalogs: vercel/mistral-large offers none/high, and xai/grok-4.3 adds xhigh on the direct xAI route.

Commits

• 25a837a release: bump action to 0.1.102
• 5bd7439 security: close credential and exfil paths found in the hardening audit (#1568)
• b49ccff Clear the Dependabot backlog: 306 alerts, both lockfiles (#1570)
• d75bd1c models: absorb Zen's new K3 and Qwen Max effort ladders, and retire the free ...
• 82247c3 models: add Zen's free Beam, hidden until a release carries it (#1551)
• e5e7589 chore(models): bump vercel/deepseek-flash to deepseek-v4.1-flash-1017 (#1530)
• ae4145b release: bump action to 0.1.101
• 795228a Cap the funded trial at $10 of credits on GPT-6 Luna, and rewrite every failu...
• a68236b Re-review when Pullfrog's own push dismisses its approval; let the gh token r...
• 9a8e9d4 release: bump action to 0.1.100
• Additional commits viewable in compare view

Dependabot will resolve any conflicts with this PR as long as you don't alter it yourself. You can also trigger a rebase manually by commenting @dependabot rebase.

---

Dependabot commands and options

You can trigger Dependabot actions by commenting on this PR:

• @dependabot rebase will rebase this PR
• @dependabot recreate will recreate this PR, overwriting any edits that have been made to it
• @dependabot show <dependency name> ignore conditions will show all of the ignore conditions of the specified dependency
• @dependabot ignore <dependency name> major version will close this group update PR and stop Dependabot creating any more for the specific dependency's major version (unless you unignore this specific dependency's major version or upgrade to it yourself)
• @dependabot ignore <dependency name> minor version will close this group update PR and stop Dependabot creating any more for the specific dependency's minor version (unless you unignore this specific dependency's minor version or upgrade to it yourself)
• @dependabot ignore <dependency name> will close this group update PR and stop Dependabot creating any more for the specific dependency (unless you unignore this specific dependency or upgrade to it yourself)
• @dependabot unignore <dependency name> will remove all of the ignore conditions of the specified dependency
• @dependabot unignore <dependency name> <ignore condition> will remove the ignore condition of the specified dependency and ignore conditions

[pullfrog]

OpenCode Zen returned an error, so this run stopped. If it names your key, plan or balance, fix that with the provider or pick another model.

This API key has expired. Create a new API key in OpenCode Console and replace it in your client or secret store. Signing in again will not renew this key.

Pick another model →

  ^{｜ Rerun failed job ➔ ｜ View workflow run ｜ via Pullfrog ｜ Using big-pickle (free)}

[kilo-code-bot]

Code Review Summary

Status: No Issues Found | Recommendation: Merge

Dependabot patch bump of pullfrog/pullfrog from 0.1.98 to 0.1.102. The updated pin 25a837a2b078177c3acdce790506d1fd1acedd82 matches the 0.1.102 release commit and remains a full-commit SHA pin, which is the recommended supply-chain practice. No workflow inputs, permissions, or secrets handling changed.

Files Reviewed (1 file)

• .github/workflows/pullfrog.yml

---

_{Reviewed by efficient · Input: 21.5K · Output: 1.1K · Cached: 102.8K}

[pullfrog]

Warning

Your OpenCode Zen API key has expired, so every run stops before the agent starts.

Every Pullfrog run on get-flashbacks has failed since October 17 (5 runs, no successes), so this review did not happen.

OpenCode Zen returns "This API key has expired. Create a new API key in OpenCode Console and replace it in your client or secret store" for the key stored on your Pullfrog account. The repo is set to use opencode/big-pickle, which requires a valid OpenCode key, and no other provider key is stored for that model.

To fix it:

1. Replace OPENCODE_API_KEY on the BYOK tab of your Billing card with a fresh key from OpenCode Console.
2. You can also run npx pullfrog auth codex in the repo to use a ChatGPT subscription instead — the setup docs have the details.