fix(dispatching-parallel-agents): align "Use when" threshold with description

[jason-h23]

What problem are you trying to solve?

The frontmatter description says "2+ independent tasks" but the "Use when" bullet list says "3+ test files failing with different root causes". An agent reading this skill gets conflicting signals about when to dispatch parallel agents — should it wait for 3 failures or 2?

What does this PR change?

Changes "3+" to "2+" in the "Use when" section to match the frontmatter description.

Is this change appropriate for the core library?

Yes — it's a one-line consistency fix in an existing core skill. No new functionality.

What alternatives did you consider?

1. Change description to "3+" instead — but the skill's purpose is dispatching for any set of independent failures, and 2 independent failures already justify parallel dispatch.
2. Leave both — but conflicting thresholds cause unpredictable agent behavior.

Does this PR contain multiple unrelated changes?

No. Single file, single line.

Existing PRs

• I have reviewed all open AND closed PRs for duplicates or prior art
• Related PRs: fix: restore worktree creation step in brainstorming skill #829 (worktree/brainstorming fix — separate issue, originally bundled in my earlier fix: minor threshold, reference, and terminology fixes across skills #798 which I've now closed)

Environment tested

Harness	Harness version	Model	Model version/ID
Claude Code	1.0.33	Claude Opus 4.6	claude-opus-4-6

Evaluation

Verified by reading the file: description field on line 3 says "2+ independent tasks", "Use when" on line 37 said "3+". After change, both say "2+".

Not a behavior-shaping change (rationalization table, red flags, etc.) — purely a number consistency fix.

Rigor

• If this is a skills change: I used superpowers:writing-skills and completed adversarial pressure testing — N/A, this is a number fix, not a workflow/behavior change
• This change was tested adversarially, not just on the happy path — verified no other references to "3+" exist in the file
• I did not modify carefully-tuned content (Red Flags table, rationalizations, "human partner" language) without extensive evals showing the change is an improvement

Human review

• A human has reviewed the COMPLETE proposed diff before submission