[EricTechPro]

Note

⏳ In review · round 1 · ✅ 5/5 AC · 56 offline suites and 13/13 hosted checks pass · head 78a4855

Problem

• Where: Setup and worker dispatch on Claude or Codex
  • a. Record a helper pack in a lockfile, or install only a few helpers
  • b. Select a card whose remaining lanes need additional skills
  • c. Dispatch the card
• Who: Super Board operators
• What happened: Setup could report helper readiness from a lockfile, while dispatch could claim cards and start workers without their required host-visible skills.

Solution

• One stdlib resolver owns ordered label chains, Build overrides, lane boundaries, conditional helpers and the two declared ponytail fallbacks.
• Resolve readable, nonempty skill files in the actual host's visible roots or enabled, scoped installation metadata. Codex discovery stops at its owning Git root and honors native disabled paths.
• Unknown native activation syntax and trust-sensitive project activation require an effective active catalog. Catalog-only resolution never falls back to inactive files or caches.
• Gate remaining chains before dispatch claims and Codex wave markers/router/children. Recheck workers and conditional helpers before activation. A skill-readiness halt preserves claims, worktrees and approval evidence and stops only the runner's workers.
• Keep explicit onboard helper repair, then re-resolve actual files. Installer success alone cannot make setup green; read-only checks and dispatch install nothing.
• Docs: https://learn.chatgpt.com/docs/build-skills — host discovery and local skill activation.
• Docs: https://learn.chatgpt.com/docs/config-file/config-basic — project configuration trust and layers.

Acceptance criteria

• Lockfiles, partial installs, broken/empty/unreadable files, wrong hosts and inactive plugin paths cannot satisfy required skills.
  ↳ tests/test_skills_readiness.py exercises real isolated files and catalogs through the public CLI.
• Exact label precedence, Build overrides, required lane boundaries and declared fallbacks produce the selected chain.
  ↳ Literal ordered policy fixtures cover bug, feature, ux, refactor, tech-debt, tests, docs, QA-only and Review.
• Missing skills halt actual launchers before workers or new claims, preserving existing evidence.
  ↳ tests/test_skill_dispatch.py covers all four entry routes, Codex fallback, legacy dispatch and standalone build.
• Files removed after initial wave readiness halt before the next worker; legacy halt stops only owned workers.
  ↳ Real Codex grade-to-lane and full legacy launcher fixtures preserve unrelated processes, locks and approval bytes.
• Setup keeps repair compatibility and reports actual readiness after successful installers.
  ↳ Public check/fix fixture proves a successful fake npx cannot falsely report helpers installed; --no-helpers preserves missing-readiness truth.

Iteration history

Lane	Done	Time	Details
🔨 builder	✅	Oct 13 PDT	historical red witnesses; focused resolver and real dispatch fixtures pass

Risk

🟡 Medium · launch-time gates now fail closed on unavailable helpers or unsupported activation metadata. Repair exact host-visible files or supply the effective active catalog and restart explicitly. Eight shipped skills, three lanes, approval and merge rules remain intact. Workflow runtime checks depend on workers obeying their first-operation instruction; the orchestrator's preclaim CLI is the deterministic gate. No paid host or live board run was used for validation.

[qodo-code-review]

Qodo reviews are paused for this user.

Troubleshooting steps vary by plan Learn more →

On a Teams plan?
Reviews resume once this user has a paid seat and their Git account is linked in Qodo.
Link Git account →

Using GitHub Enterprise Server, GitLab Self-Managed, or Bitbucket Data Center?
These require an Enterprise plan - Contact us
Contact us →