Hacktoberfest 2026: the issues maintainers tagged for October, open and beginner-friendly. Browse Hacktoberfest issues

[aw] Prompt Optimization reported incomplete result

Open
  • Is something wrong?
#309 1 comment 0 reactions 0 assignees View on GitHub

Maintainers usually reply within 1 day

Nobody has claimed this yet.

Assessment

Difficulty
4/5
Estimated time
3-5 days
Newbie friendliness
38/100
Issue type
Bug
Clarity
Mostly clear
Activity status
Active
Tech stack
docker, github-actions, javascript
Domain
ai-infra-agents, ci-cd, infrastructure

Research direction

Start with .github/workflows/prompt-optimization.md and the agentic-workflows skill, then inspect scripts/prompt-optimizer.mjs and the failed run logs. Trace how the Ollama evaluation server is restored, served, and exposed to the agent container. Done means the workflow can reach the configured models and complete the prompt-optimization evaluation on a rerun.

Written by the indexing model from the issue text.

Description

agentic-workflows
Workflow Failure

Workflow: Prompt Optimization
Branch: main
Run: https://github.com/githubnext/gh-aw-wizard/actions/runs/37157539293

[!WARNING]
Task Could Not Be Completed: The agent reported that the task could not be performed due to an infrastructure or tool failure.

Reasons:

  • Ollama eval server at (127.0.0.1/redacted) (host-reflected via --enable-host-access --allow-host-ports 11434) is unreachable from the agent sandbox, so no scenario-prompt evaluation or hill-climbing could be run.

    No open PR titled "Prompt optimization:" exists (guard check via gh pr list confirmed only #304, unrelated), so the task proceeded to the hill-climbing loop.

Connectivity diagnosis:

  • curl (127.0.0.1/redacted) and /v1/models`: connection refused.
  • No ollama process, no ~/.ollama cache directory, and no .ollama/GGUF blob files found anywhere on the filesystem searchable from this sandbox — the model-restore/serve steps appear to have run in the GitHub Actions host context, not inside this agent container.
  • Tried all plausible host-access paths: host.docker.internal (DNS SERVFAIL via the container's embedded resolver 127.0.0.11), Docker bridge gateway candidates 172.30.0.1 and 172.17.0.1 (connection refused/timeout), direct squid-proxy CONNECT tunnel to 172.17.0.1:11434 (squid reported "Connection to 172.17.0.1 failed... Connection refused"), squid HTTP-proxy to 172.17.0.1:11434 (403 Access Denied — not in squid's allowlist), and the api-proxy container IP on ports 3128/8080/8000 (connection refused).
  • /etc/hosts inside the agent container only lists the firewall sidecar containers (squid-proxy, api-proxy, cli-proxy, awmg-mcpg/cli-proxy) — no host-gateway entry is present despite the docker-compose audit file showing extra_hosts: host.docker.internal: host-gateway on the squid-proxy and cli-proxy services (not on the agent service itself).
  • No sudo/root capability available to inspect iptables NAT rules or network namespaces further (blocked by no-new-privileges, and inspecting such internals is disallowed by policy regardless).

Net effect: scripts/prompt-optimizer.mjs --evaluate --all-models --eval-url (127.0.0.1/redacted) ... cannot reach either the Qwen2.5-1.5B or SmolLM2-360M proxy model, so no incumbent score, candidate, or hill-climbing step could be measured. No repository files were changed (a pre-existing unrelated modification to .github/skills/optimize-scenario-prompt/SKILL.md was reverted). No PR, review comment, or review was created.

This is a structured incompletion signal (report_incomplete), not a real task outcome. Any other safe outputs emitted alongside this signal (e.g., comments) describe the failure state, not a completed review or action.

Action Required

Assign this issue to an agent to debug and fix the issue.

Debug with any coding agent

Use this prompt with any coding agent (GitHub Copilot, Claude, Gemini, etc.):

Debug the agentic workflow failure using https://raw.githubusercontent.com/github/gh-aw/main/debug.md

The failed workflow run is at https://github.com/githubnext/gh-aw-wizard/actions/runs/37157539293
Manually invoke the agent

Debug this workflow failure using your favorite Agent CLI and the agentic-workflows prompt.

[!TIP]

Stop reporting this workflow as a failure

To stop a workflow from creating failure issues, set report-failure-as-issue: false in its frontmatter:

safe-outputs:
  report-failure-as-issue: false

Generated from Prompt Optimization · copilot · 152.8 AIC · ◷

Dominant language
JavaScript
Stars
6
Forks
1
Avg merge
1d 5h
Merged PRs (30d)
30

Getting set up

This project ships no dev container, Dockerfile or contributing guide, so setting up is up to you: start from its README, and see our first-contribution guide for the general steps.

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

More from githubnext/gh-aw-wizard

All issues in githubnext/gh-aw-wizard

Similar issues

More JavaScript issues

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.