LOW findings from the independent review of #2366/#2367/#2369, tracked instead of fixed in-round (law 2). (1) docs/GITHUB_PROJECT_AUTOMATION.md line ~156 still describes the scripts/github/* helpers as fallbacks for GitHub MCP gaps; with GitHub MCP removed from .mcp.json they are Claude's primary path - runtime-qualify the sentence. (2) .mcp.json pins chrome-devtools-mcp@1.8.0 while .codex/config.toml still uses @latest; either pin both or say in docs/MCP_TOOLING_GUIDE.md section 0.1 that the DevTools version is per-runtime. (3) Optional: an executable negative control for the unwind path in Test-Assert-TaskdeckCheckoutFingerprint.ps1 replacing the removed AFTER-TRY-RAN marker. Refs #1291, #1138.
LOW findings from the independent review of #2366/#2367/#2369, tracked instead of fixed in-round (law 2). (1) docs/GITHUB_PROJECT_AUTOMATION.md line ~156 still describes the scripts/github/* helpers as fallbacks for GitHub MCP gaps; with GitHub MCP removed from .mcp.json they are Claude's primary path - runtime-qualify the sentence. (2) .mcp.json pins chrome-devtools-mcp@1.8.0 while .codex/config.toml still uses @latest; either pin both or say in docs/MCP_TOOLING_GUIDE.md section 0.1 that the DevTools version is per-runtime. (3) Optional: an executable negative control for the unwind path in Test-Assert-TaskdeckCheckoutFingerprint.ps1 replacing the removed AFTER-TRY-RAN marker. Refs #1291, #1138.