[aw-failures] [aw] Fix aw-portfolio-yield: placeholder npm package + empty OTel endpoint break compile and runtime

[github-actions]

Problem

.github/workflows/aw-portfolio-yield.md (introduced by #31363 on 2026-05-11 ~05:35 UTC) imports .github/workflows/shared/otel-observability.md, which contains:

• a placeholder npm package @your-org/otel-query-mcp that does not exist on the npm registry; and
• an unconditional gateway.opentelemetry.endpoint set to ${{ secrets.OTLP_ENDPOINT }}, which resolves to the empty string in this repo (the secret is unset).

This causes two separate downstream failures:

1. Compile-time: Every run of the Agentic Maintenance workflow's compile-workflows job now fails with:

   .github/workflows/aw-portfolio-yield.md:1:1: error: runtime package validation failed
     npx package '`@your-org/otel-query-mcp`' not found on npm registry: npm error code E404
   ✗ compilation failed
   ##[error]Process completed with exit code 1.
   

   Reproduced in §25657460521 and §25654651003; the last green Agentic Maintenance run §25653034240 was at 05:58 UTC, 5 minutes before Add Agentic Workflow Portfolio Yield workflow #31363 was pushed.

2. Runtime: Agentic Workflow Portfolio Yield itself can't start because the MCP Gateway (v0.3.6) rejects its config:

   config:validation_schema Schema validation failed:
     jsonschema: '/gateway/opentelemetry/endpoint' does not validate with .../endpoint/minLength: length must be >= 1, but got 0
     Error: does not match pattern '^((redacted)+|\${[A-Za-z_][A-Za-z0-9_]*})$'
   failed to load config: Configuration validation error (MCP Gateway version: v0.3.6)
   ##[error]Process completed with exit code 1.
   

   Reproduced in §25654663141. Auto-issue [aw] Agentic Workflow Portfolio Yield failed #31439 tracks the symptom; this issue tracks the root cause and fix.

Affected workflows

• Agentic Maintenance — entire compile-workflows job and everything downstream of it is now skipped on every schedule. This is high-blast-radius: maintenance includes cleanup-cache-memory, close-expired-entities, update_pull_request_branches, apply_safe_outputs, close_agentic_workflows_issues — all currently not running.
• Agentic Workflow Portfolio Yield — 100% failure rate (can't even reach the agent step).
• Any other workflow that imports shared/otel-observability.md is at the same risk.

Affected run IDs

• §25657460521 — Agentic Maintenance (schedule, 07:52 UTC)
• §25654651003 — Agentic Maintenance (push, 06:44 UTC)
• §25654663141 — Agentic Workflow Portfolio Yield (workflow_dispatch, 06:44 UTC)

Probable root cause

PR #31363 was authored against a template/example version of shared/otel-observability.md whose values (@your-org/otel-query-mcp, hard OTLP_ENDPOINT reference) were never replaced with real ones before merge. Compile-time package validation was bypassed because the new file didn't itself fail compile — aw-portfolio-yield.md's lock file was already present in the merge commit, and the compile-workflows regeneration didn't run until the next scheduled Agentic Maintenance.

Proposed remediation

Pick whichever scope matches intent:

1. Minimum fix (P0) — Edit .github/workflows/shared/otel-observability.md:

   • Remove the mcp-servers.otel block, or replace @your-org/otel-query-mcp with the real published MCP server package name.
   • Wrap or remove the observability.otlp.endpoint block so the gateway config is omitted when OTLP_ENDPOINT is unset (or write a literal (redacted) URL).
   • Run gh aw compile .github/workflows/aw-portfolio-yield.md locally and confirm 0 errors before committing.

2. Defense in depth (P1) — Add a CI guard so any new shared file referencing an unpublished npm package is rejected at PR time, not at the next scheduled Agentic Maintenance. The Agentic Maintenance compile-workflows job already does this validation; gating PR merges on it would have caught Add Agentic Workflow Portfolio Yield workflow #31363 before merge.

Success criteria / verification

• gh aw compile .github/workflows/aw-portfolio-yield.md exits 0 with no runtime.packages errors.
• gh aw compile over the whole repo produces ✗ Compiled N workflow(s): 0 error(s) again.
• A fresh workflow_dispatch of Agentic Workflow Portfolio Yield advances past the MCP Gateway startup step (gateway logs do not contain Configuration validation error).
• The next scheduled Agentic Maintenance run reports compile-workflows success.

References

• Source: .github/workflows/aw-portfolio-yield.md (introduced by Add Agentic Workflow Portfolio Yield workflow #31363)
• Imported: .github/workflows/shared/otel-observability.md
• Parent failure report: see linked aw_parent01
• Symptom auto-issue: [aw] Agentic Workflow Portfolio Yield failed #31439

Related to #30961

Generated by [aw] Failure Investigator (6h) · ● 39.5M · ◷

• expires on May 18, 2026, 8:17 AM UTC