validating

name	validating
description	Pre-execution validation gate — structural, scope, and readiness checks before workers run

The Validating skill is the mandatory gate between planning and execution. It catches structural errors in task definitions, file conflicts between parallel tasks, and unresolved risks before any Worker touches the codebase. A BLOCKED result stops execution until issues are resolved.

<Use_When>

• After the Architect produces task files (Phase 4 of start link workflow)
• Before spawning any Workers — sequential or parallel
• When the Master needs to confirm the plan is safe to execute </Use_When>

<Do_Not_Use_When>

• Start Fast workflow (no formal task files exist)
• Re-running after a Worker completes (use Reviewer instead)
• Already ran and returned READY for the current plan version </Do_Not_Use_When>

Pass 1 — Structural Verification

Read all task JSON files from .oh-my-link/tasks/.

For each task, verify:

• link_id present and unique across all tasks
• title present and non-empty
• description present and non-empty
• acceptance_criteria is a non-empty array
• file_scope is a non-empty array
• status is "pending"
• depends_on field contains only valid link_id values

Dependency graph checks:

• No cycles (DFS with visited + in-stack detection)
• No self-referencing link_id in its own depends_on list

Fail conditions → add to BLOCKED list with specific task ID and field.

Pass 2 — Scope Verification

For each file_scope entry across all tasks:

1. File existence: check whether the file exists on disk or is explicitly marked as a new file to be created. Unknown paths are flagged as warnings (not blockers unless ambiguous).

2. Conflict detection: build a map of file_path → [link_ids]. Any file appearing in two or more tasks that could run concurrently is a CONFLICT. Concurrent = tasks with no dependency relationship between them.

3. Scale check: if total unique file count exceeds 30, flag for manual review. This is a warning, not a blocker.

Pass 3 — Readiness Assessment

Review the plan holistically:

• Are there any unresolved questions from the Scout phase (check .oh-my-link/plans/decisions.md if present)?
• Are there tasks marked HIGH risk without a spike or prototype task preceding them?
• Are acceptance criteria testable and specific (not vague like "works correctly")?

Flag issues that would predictably cause Worker failures.

Output

If all checks pass → READY

VALIDATION: READY

Structural: PASS (N tasks, no cycles, all fields present)
Scope: PASS (N files, no conflicts)
Readiness: PASS

Proceed to Gate 3 — choose execution mode:
  [S] Sequential  — one Worker at a time, safest
  [P] Parallel    — concurrent Workers on non-overlapping tracks

Do NOT spawn any Workers until the user responds to Gate 3. Wait for explicit user selection of Sequential or Parallel.

If any check fails → BLOCKED

VALIDATION: BLOCKED

Issues found:
  1. [STRUCTURAL] task-03: acceptance_criteria is empty
  2. [CONFLICT] file src/auth/middleware.ts claimed by task-01 and task-04
  3. [READINESS] task-06 is HIGH risk with no preceding spike

Resolve these issues before execution can proceed.

<Tool_Usage>

• Read: load task JSON files from .oh-my-link/tasks/
• Glob: enumerate task files
• Bash: check file existence for scope verification
• No Agent spawning — Validating is a leaf agent
• Write: optionally produce .oh-my-link/plans/validation-report.md </Tool_Usage>