Troubleshooting

No review was posted

Work through this checklist from the top — one of these is almost always the culprit:

1. Did you trigger a review? Comment /pura review as the first line of a PR comment — or open the PR, push new commits, or mark it ready for review, which all start a review automatically.
2. Is the PR a draft? Drafts are skipped — mark it ready for review and PURA picks it up automatically.
3. Is the PR ignored? A repo admin may have disabled PURA for the PR with a bare /pura ignore comment — auto-reviews pause on that PR. A manual /pura review still works and reactivates it.
4. Is PURA installed on that repository? Check the GitHub App's repository access. You may have installed it on only certain repos, and this one might not be in the list.
5. Is your account active? An expired trial with no paid plan pauses reviews until you pick a plan.
6. Is the monthly review allowance used up? Each plan includes a set number of reviews per month. Once it's reached, PURA pauses reviews (with a note on the PR) until the 1st of next month — check usage in the dashboard or upgrade your plan.
7. Is the repository too large? Each plan limits the download size of the repository at the review commit (Starter: 25MB, Scale: 50MB, Pro: 100MB). If your repo exceeds that limit, PURA skips the review with a note on the PR. Upgrade your plan for a higher limit.

Budget reached

If an agent has hit its daily, weekly, or monthly cap — or a scope cap on this PR's team, repo, author, or labels — and no other provider in the chain fits, PURA pauses the review rather than overspending. You have a few options:

• Raise the cap — the agent's overall cap, or the scope cap that covers this PR, in the dashboard.
• Add another provider to the agent so PURA has a fallback.
• Wait for the window to reset — daily caps reset overnight, weekly caps reset on Monday, and monthly caps reset on the 1st.

No agent matched the PR

A review only runs if there's an Active review agent, with a working provider key, whose scope allows that PR. A few things to check:

• Is the agent set to Active? Inactive agents are never used.
• Does the agent's first provider have a key pasted in? Agents without a key on their first provider are never selected. If a review does run but errors at review time, the key itself is usually wrong — double-check it in the dashboard.
• Does the agent's routing allow this PR? The agent's Review Skill (.agents/skills/<agent-name>/SKILL.md) may carry .pura-routing instructions that skip certain changes — a rule like “do not run for docs-only changes” means this agent won't review docs PRs. If every eligible agent skips, no review is posted.
• If you've scoped agents by team, repo, author, or PR label, make sure at least one of them allows this particular PR. If none match, PURA has nowhere to route the review.

It used a model I didn't expect

Each provider on an agent has three model slots — default, high, and low — and the agent's routing instructions pick the tier per PR. If the result surprises you, you have two levers:

• Adjust the agent's routing instructions — refine the .pura-routing rules in the agent's Review Skill so the right changes select the right tier.
• Set the models on the agent — put the models you want in the default, high, and low slots.

---
name: security-agent
description: Security-focused code review.
.pura-routing: |
  Use this only for nodejs files. For large changes, use the high model.
---

Review for security issues: injection, secrets, and auth mistakes.

Write plain English intent — PURA evaluates these rules per PR against the changed files and description, and selects the matching model tier.

Still stuck?