fix(skills): require user confirmation before skill updates

[santhiprakash]

What

Replace the "run silently, don't ask" skill-freshness directive with explicit user-confirmation guidance in ten workflow SKILL.md files, and regenerate skills-manifest.json.

Why

Fixes the trust-model issue in #2613: published skills were instructing agents to run npx hyperframes skills update without surfacing it to the user. skills/general-video/SKILL.md already used safer wording; this aligns the remaining workflow skills.

How

One-line wording change per affected skill:

• Before: run silently, don't ask
• After: confirm with the user before running

No CLI or runtime behavior changes.

Test plan

• bun packages/cli/scripts/gen-skills-manifest.ts --check — manifest in sync
• rg "run silently, don't ask" skills -g SKILL.md — zero matches
• rg "confirm with the user before running" skills -g SKILL.md — 10 matches (workflow skills)
• CI: Skills manifest + docs checks

Notes

• Prior docs(skills): require confirmation before skill updates #2948 was closed to keep unrelated docs in docs: fix preview port in DOCS_GUIDELINES example #2903; this is a focused follow-up on current main.
• Issue Global skill-mirroring step deletes unrelated user-authored skills, then erases evidence of doing so #3111 / PR fix(cli): stop skills update deleting skills the manifest never covered #3118 are unrelated (data-loss fix); not duplicated here.

Fixes #2613 (silent self-update portion only; shell-injection and SRI items remain open in the issue).