refactor: remove automated review response workflow

[kolatts]

Superseded by #293, which is based cleanly on the squash-merged #290 and contains only the intended workflow deletion and Imagile Bot review-gate update.

[github-actions]

Claude finished @kolatts's task in 2m 21s —— View job

---

PR Review

• Clean up previous Claude comments (no prior comments found)
• Read changed workflow files
• Analyze changes and provide review
• Submit formal GitHub review — Approved

---

claude-triage.yml: The Imagile Bot token generation is correctly structured. Minimal job-level permissions (contents: read) with elevated privileges flowing only through the explicit app token is the right pattern. Bot identity resolution (app-slug + [bot] suffix) is solid, and bot_id/bot_name correctly establish the identity context for the Claude action.

The CRITICAL SAFETY RULE block is well-placed before the untrusted issue body, and the === BEGIN/END === delimiters give the LLM a clear trust-boundary signal. The --allowedTools in claude_args is appropriately narrow.

Belt-and-suspenders label removal (both in the prompt step 5 and the workflow's "Remove triggering label" step) is intentional and fine.

claude-review.yml: The branch pattern gate (startsWith(..., 'claude/issue-')) correctly scopes which Imagile Bot PRs get reviewed. The claude[bot] allowance without a branch restriction is acceptable given the upstream fork guard already blocks external actors. Narrowing allowed_bots to imagile-bot reduces the attack surface appropriately.

The workflow-change token fallback (using github.token when claude-review.yml itself changes) is a practical workaround for the Claude app's default-branch validation behavior.

Deletion of claude-respond.yml: No issues — removing a workflow that was skipping or failing on nearly every run is the right call.

[github-actions]

Clean refactor — good security hygiene throughout.

claude-triage.yml: The Imagile Bot token generation is correctly structured. Minimal job-level permissions (contents: read) with elevated privileges flowing only through the explicit app token is the right pattern. The bot identity resolution (app-slug + [bot] suffix) is solid, and the bot_id/bot_name parameters correctly establish the identity context for the Claude action.

The CRITICAL SAFETY RULE block is well-placed before the untrusted issue body, and the === BEGIN/END === delimiters give the LLM a clear signal of the trust boundary. The --allowedTools in claude_args is appropriately narrow.

Belt-and-suspenders label removal (in both the prompt step 5 and the workflow's "Remove triggering label" step) is intentional and fine.

claude-review.yml: The branch pattern gate (startsWith(..., 'claude/issue-')) correctly scopes which Imagile Bot PRs get reviewed. The claude[bot] allowance without a branch restriction is acceptable given the upstream fork guard. Narrowing allowed_bots to imagile-bot is the right call.

The workflow-change token fallback (using github.token when claude-review.yml itself changes) is a practical workaround for the Claude app's default-branch validation behavior.

Deletion of claude-respond.yml: No objections — removing a workflow that was skipping or failing on nearly every run is unambiguously the right call.