fix(skills): disambiguate 'grade' triggers, trim loop-evals description, qualify loop-run example

[SollanSystems]

Closes #34.

Three frontmatter-only trigger fixes — no body or behavior changes:

• loop-evals: the bare-"grade" trigger is now noun-qualified — "or grade a run's outcome against its SPEC" — and the description is trimmed from 736 to 526 chars (sibling band is 398–541) by dropping the capability-summary sentence whose five clauses each map 1:1 to an existing body H2 (7-layer suite, FCR/RP first-class metrics, deterministic-first-then-rubric, judge calibration, repo-native regression harness).
• loop-inspector: "grade a superpowers / ruflo / .loop harness" → "grade this harness/contract's readiness (a superpowers / ruflo / .loop layout)" — noun-first, all three harness-name trigger tokens retained (541 chars).
• loop-run: first example qualified — 'run the loop' → 'run the agent loop' (430 chars).

One deliberate deviation from the issue's literal drop range: removing both trailing sentences lands at 340 chars, below the 398 floor — so the delegation/"/verify-slice" sentence (the one the issue itself flags as load-bearing disambiguation) stays in the description verbatim.

Gates (run twice — implementer + independent verifier):

• validate_frontmatter.py — 9 skills, 0 errors
• self_eval.py — 13/13
• full suite — 393 passed / 14 skipped (skips are the langgraph/temporalio env guards; those run in CI)

All descriptions remain quoted YAML scalars. No bare-"grade" collision remains.

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.