fix(github_action): skip bot-sender comments to avoid feedback loop

[IsmaelMartinez]

PR Summary by Qodo

Skip bot-authored GitHub Action comments to prevent issue_comment feedback loops
🐞 Bug fix 🧪 Tests 🕐 10-20 Minutes

Walkthroughs

User Description

What

When pr-agent runs as a GitHub Action on issue_comment events, it re-fires on the comments it posts itself (e.g. "Preparing review..."). Each bot comment triggers another issue_comment run, which parses the comment body as a command and fails with Unknown command: preparing, as described in #2398.

Fix

Skip processing in the issue_comment / pull_request_review_comment handler when the comment's sender is a Bot. This mirrors the if: github.event.sender.type != 'Bot' guard the docs already recommend, so users no longer have to add it to every workflow themselves.

Behaviour note: this skips all bot-authored comments. That matches the documented sender.type != 'Bot' recommendation; if someone intentionally drives pr-agent from another bot, they'd need a different trigger. Happy to gate it behind a config flag instead if you'd prefer.

Tests

• test_issue_comment_from_bot_sender_is_skipped — a bot-authored comment is not handled.
• test_issue_comment_from_user_is_processed — a human comment is still handled (the guard doesn't over-skip).

Fixes #2398.

AI Description

• Skip issue_comment processing when the comment sender is a GitHub Bot
• Prevent pr-agent from re-triggering on its own status comments and failing command parsing
• Add regression tests ensuring bot comments are skipped while human comments still run

Diagram

graph TD
  A{{"GitHub comment event"}} --> B["github_action_runner.run_action"] --> C{"sender.type == Bot?"}
  C -- "yes" --> D["Log + return"]
  C -- "no" --> E["Parse body + URL"] --> F["PRAgent.handle_request"] --> G["GitHub API"]
  subgraph Legend
    direction LR
    _evt{{"Event"}} ~~~ _handler["Handler"] ~~~ _dec{"Decision"}
  end

Loading

High-Level Assessment

The following are alternative approaches to this PR:

1. Workflow-level guard only (docs approach)

• ➕ Keeps library behavior unchanged
• ➕ Allows users to decide whether other bots should be able to trigger pr-agent
• ➖ Easy to miss; every workflow must remember to add it
• ➖ Doesn’t help users already experiencing the loop until they change workflows

2. Config flag (e.g., skip_bot_senders=true by default)

• ➕ Maintains safe default while allowing opt-out for bot-driven automations
• ➕ Makes behavior explicit and discoverable in configuration
• ➖ Adds configuration surface area and support burden
• ➖ Requires documenting and threading config through the action runner

3. Allowlist/denylist by sender.login (skip only pr-agent bot)

• ➕ Preserves ability to trigger from other bots
• ➕ More targeted than sender.type == Bot
• ➖ Brittle across installations/forks/renamed apps
• ➖ Still risks loops if other bot comments trigger self-replies

Recommendation: The current sender.type == 'Bot' short-circuit is a solid default because it matches GitHub’s documented workflow guard and prevents a high-impact feedback loop out of the box. If there is a real use case for bot-driven commands, consider a follow-up that adds a config toggle or allowlist; otherwise keep the minimal guard for safety and simplicity.

File Changes

Bug fix (1)

github_action_runner.py Skip bot-authored comment events in GitHub Action runner +8/-0

Skip bot-authored comment events in GitHub Action runner

• Adds an early-return guard in the issue_comment / pull_request_review_comment handler when the event sender is a Bot. This prevents pr-agent from re-triggering on its own comments and entering a command-parsing feedback loop (issue #2398).

pr_agent/servers/github_action_runner.py

---

Tests (1)

test_github_action_runner_core.py Add regression tests for bot-sender skip behavior +83/-0

Add regression tests for bot-sender skip behavior

• Introduces fixtures/helpers to isolate global settings mutations and to stub dependencies for issue_comment handling. Adds async tests verifying bot-authored comments are skipped and human comments still invoke PRAgent handling.

tests/unittest/test_github_action_runner_core.py

---

[qodo-free-for-open-source-projects]

Code Review by Qodo

🐞 Bugs (0) 📘 Rule violations (0) 📎 Requirement gaps (0)

Great, no issues found!

Qodo reviewed your code and found no material issues that require review