fix(nighthawk): prevent pkill self-termination via variable indirection

[Copilot]

The Nightly Issue Implementer was killing itself with exit code 143 (SIGTERM) on every run that reached the cleanup step. Root cause: gh-aw passes the full prompt as a CLI argument to the claude process — so pkill -f "next start" matched the agent's own /proc/PID/cmdline (which contained the literal string next start from the prompt instructions).

Changes

• Shell variable indirection for all server-stop commands — replaces the literal pattern with a variable so it never appears verbatim in the prompt text:

  _SRV=start
  pkill -TERM -f "next $_SRV" 2>/dev/null || true
  sleep 2
  pkill -KILL -f "next $_SRV" 2>/dev/null || true

  At runtime bash expands $_SRV → start; but the string next start is absent from the prompt, so pkill no longer matches the agent process.

• Callout updated (Step 8 bash gateway constraints) — explains the self-match mechanism explicitly and mandates the variable indirection pattern going forward.

• Timeout reference corrected — Step 10 referenced timeout-minutes: 60; changed to timeout-minutes: 45 to match the actual workflow frontmatter.

No lock.yml changes needed — the prompt is loaded at runtime via {{#runtime-import .github/workflows/nighthawk.md}}.

Warning

Firewall rules blocked me from connecting to one or more addresses (expand for details)

I tried to connect to the following addresses, but was blocked by firewall rules:

• https://api.github.com/repos/unic/ai-developer-hub/actions/runs/25317528503
  • Triggering command: /usr/bin/gh gh api repos/{owner}/{repo}/actions/runs/25317528503 --jq {databaseId: .id, number: .run_number, url: .html_url, status: .status, conclusion: .conclusion, workflowName: .name, workflowPath: .path, createdAt: .created_at, startedAt: .run_started_at, updatedAt: .updated_at, event: .event, headBranch: .head_branch, ) if 'safeoutputs' in str(data).lower() or 'create_pull_request' in str(data).lower() or 'add_comment' in str(data).lower(): print(json.dumps(data)[:300]) -j DROP (http block)
• https://api.github.com/repos/unic/ai-developer-hub/actions/workflows
  • Triggering command: /usr/bin/gh gh workflow list --all --json id,name,path,state (http block)

If you need me to access, download, or install something from one of these locations, you can either:

• Configure Actions setup steps to set up my environment, which run before the firewall is enabled
• Add the appropriate URLs or hosts to the custom allowlist in this repository's Copilot coding agent settings (admins only)

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
ai-developer-hub	Ready	Preview, Comment	May 5, 2026 10:45am

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you all sign our Contributor License Agreement before we can accept your contribution.
1 out of 2 committers have signed the CLA.

✅ studert
❌ Copilot
_{You have signed the CLA already but the status is still pending? Let us recheck it.}

[Copilot]

Pull request overview

This PR updates the Nighthawk workflow prompt documentation to prevent the agent from self-terminating during cleanup when stopping the local Next.js server. It addresses a real failure mode where pkill -f could match the agent process itself because the workflow prompt is passed as part of the agent process command line.

Changes:

• Replaces literal pkill -f "next start" guidance with a variable-indirection pattern (_SRV=start, pkill ... "next $_SRV") to avoid self-matching.
• Expands the Step 8 bash gateway callout to explicitly document the self-termination mechanism and the required mitigation pattern.
• Corrects the iteration timeout reference from timeout-minutes: 60 to timeout-minutes: 45 to match the workflow frontmatter.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.