Skip to content

feat(claude): allow the claude bot to trigger AI reviews#33

Closed
sfreudenthaler wants to merge 1 commit into
mainfrom
feature/allow-claude-bot
Closed

feat(claude): allow the claude bot to trigger AI reviews#33
sfreudenthaler wants to merge 1 commit into
mainfrom
feature/allow-claude-bot

Conversation

@sfreudenthaler

Copy link
Copy Markdown
Member

Problem

Bot-initiated auto-reviews fail at the anthropics/claude-code-action@v1 gate with:

Action failed with error: Workflow initiated by non-human actor: claude (type: Bot). Add bot to allowed_bots list or use '*' to allow all bots.

Example: dotCMS/infrastructure-as-code run 27365580499 — a PR whose triggering push came from the Claude Code GitHub App (actor claude, type Bot). Since Claude is our daily driver, these runs should be reviewed, not blocked.

allowed_bots is a dedicated input to claude-code-action (not a claude_args CLI flag), so it was previously unreachable through the orchestrator/executor and defaulted to "no bots allowed."

Change

  • Add an allowed_bots input to claude-orchestrator.yml and claude-executor.yml.
  • Pass it to both claude-code-action steps in the executor (Anthropic API + Bedrock paths).
  • Default it to claude org-wide, so the Claude Code GitHub App can trigger reviews on every consumer repo without per-repo config.
  • Consumers can override: allowed_bots: "" to block all bots, or allowed_bots: "*" to allow any (e.g. dependabot, renovate).

Anthropic path only — the bedrock-generic executor does not use claude-code-action and has no such gate.

Rollout

Consumers pinned at @v3.0.0 pick this up after a new tag (e.g. v3.1.0) is cut and their pin is bumped. A follow-up PR bumps the pin in dotCMS/infrastructure-as-code.

Validation

  • yaml.safe_load passes on both workflows.
  • actionlint (rhysd/actionlint:1.7.7) clean.

🤖 Generated with Claude Code

claude-code-action blocks workflow runs initiated by bot actors unless
they are in allowed_bots, which was unset. Bot-initiated PRs/pushes from
the Claude Code GitHub App (actor "claude", type Bot) therefore failed
auto-review with "Workflow initiated by non-human actor".

Expose an allowed_bots input on the orchestrator and executor and pass it
to both claude-code-action steps (Anthropic API + Bedrock). Default it to
"claude" org-wide so the Claude Code GitHub App can trigger reviews;
consumers can pass "" to block all bots or "*" to allow any.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
@sfreudenthaler sfreudenthaler requested review from a team as code owners June 11, 2026 18:14
@chatgpt-codex-connector

Copy link
Copy Markdown

You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard.

@sfreudenthaler

Copy link
Copy Markdown
Member Author

Closing without merge — decided this isn't worth changing right now. The auto-review block on bot-initiated runs is an acceptable edge case; the current GITHUB_TOKEN + read-only setup keeps things safe. Can revisit later (see #32).

@sfreudenthaler sfreudenthaler deleted the feature/allow-claude-bot branch June 11, 2026 18:30
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant