[awf] copilot-engine: cached-CLI activation skips /usr/local/bin wrapper causing ENOENT

[lpcox]

Problem
When the Copilot CLI toolcache has a valid cached entry, install_copilot_cli.sh only exports the cache dir to PATH/GITHUB_PATH and skips installing the /usr/local/bin/copilot wrapper, so the AWF harness (which hardcodes that path) cannot spawn the agent and the run fails immediately (failureClass=no_output), commonly misread as an engine/network problem.

Context
Original report: github/gh-aw#51843 (see also duplicate github/gh-aw#51373)

Root Cause
activate_cached_copilot_bin() in actions/setup/sh/install_copilot_cli.sh (lines ~481-507) returns early once GITHUB_PATH is set -- true for every GitHub Actions run -- before reaching the install -m 0755 "$wrapper_path" "${INSTALL_DIR}/copilot" step. pkg/constants/constants.go hardcodes CopilotBinaryPath = "/usr/local/bin/copilot", so only a cold/fresh install (which always installs to INSTALL_DIR) satisfies that path; a warm cache never does.

Proposed Solution
Have the AWF agent entrypoint verify /usr/local/bin/copilot exists at container start; if missing, fall back to resolving copilot via PATH/GITHUB_PATH entries and create the expected symlink/wrapper before the harness pre-flight check runs. This decouples AWF's hardcoded spawn path from upstream toolcache activation behavior.

Generated by Firewall Issue Dispatcher · auto · 30.5 AIC · ⊞ 9K · ◷

[github-actions]

👋 Hello! This issue looks like a likely duplicate of an existing open issue:

• [awf] copilot-cli install: cached activation skips /usr/local/bin wrapper, causing ENOENT #7230 - "[awf] copilot-cli install: cached activation skips /usr/local/bin wrapper, causing ENOENT"

Both issues describe the same root cause: activate_cached_copilot_bin() in install_copilot_cli.sh returns early on a cache hit (since GITHUB_PATH is always set in Actions) without installing the /usr/local/bin/copilot wrapper, and AWF's hardcoded CopilotBinaryPath then fails with ENOENT. Both reference the same upstream reports (gh-aw#51843 / #51373) and propose the same fix approaches.

If this is indeed the same underlying issue, please consider closing this one as a duplicate of #7230. Otherwise, feel free to clarify how it differs (e.g., different trigger path or failure signature).

This is an automated message from the issue duplication detector.

Generated by Issue Duplication Detector for #7241 · auto · 23.5 AIC · ⊞ 10.8K · ◷