feat(run-workflow): add run-workflow and run-featured-workflow skills

[thagarwal-sjc]

Description

Ports the run-workflow and run-featured-workflow skills from the run-workflow MCP server repo into a new creative-cloud/run-workflow plugin, adapted to this repo's frontmatter/scaffolding conventions (license field, .claude-plugin/plugin.json, .mcp.json, .tessl-plugin/plugin.json, CODEOWNERS).

run-workflow/SKILL.md is split into a slim always-loaded core plus references/ (featured-and-templates, compose, diagnostics, asset-input, alert-rca) to stay under Tessl's recommended SKILL.md token budget.

Eval scenarios cover the bare-ID inspect rule, list_actions discovery, and featured workflow routing.

Supersedes #268, which was opened from a personal fork before I had direct write access to this repo — moving it here so CI checks (Tessl eval) aren't gated behind the fork-PR reviewer-approval step.

Related Issue

No existing GitHub issue for this — happy to open one first if that's required before merge.

Motivation and Context

Brings the run-workflow MCP skills (Adobe Firefly workflow discovery/compose/run/publish, plus predefined featured workflows) into the shared Adobe skills repo so other agents/users can install them the same way as the other creative-cloud skills.

How Has This Been Tested?

• npm run validate (skills-ref spec check) — both skills valid.
• Local replication of the license-field and codeowners-coverage CI jobs — both pass.
• tessl skill lint — plugin valid, no warnings (SKILL.md token count is under budget after the references/ split).
• tessl review run (quality/security review): run-workflow scored 89%, run-featured-workflow scored 99% — both above the 50% CONTRIBUTING threshold.
• tessl eval run (real agent eval, baseline vs. with-skill): average score baseline 39% → with-skill 82%. Two scenarios (bare-ID inspect, discovery-before-answering) go from baseline 0-20% to skill 100% — clear failure-mode fixes. The third scenario (featured-workflow-routing) scores lower with-skill (47%) than baseline (97%) — a known eval-sandbox limitation, not a skill defect: the with-skill run correctly attempted real MCP tool calls and hit 403s (no live IMS credential in the sandbox), while the unguided baseline "solved" it by writing a self-contained mock implementation instead of calling any real tool. Documented inline in evals/featured-workflow-routing/task.md under "Known limitation".

Screenshots (if appropriate):

N/A — skill/documentation content only, no UI changes.

Types of changes

• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to change)

Checklist:

• I have signed the Adobe Open Source CLA.
• My code follows the code style of this project.
• My change requires a change to the documentation.
• I have updated the documentation accordingly.
• I have read the CONTRIBUTING document.
• I have added tests to cover my changes.
• All new and existing tests passed.

[github-actions]

Tessl Plugin Lint

✅ run-workflow — clean

---

✅ All 1 plugin(s) clean.

Updated by tessl-lint for commit 3169831.

[trieloff]

Seems ok to me, but I'd like to get @solimant's eyes on this, too.

[solimant]

Looks fine, but I don't see an entry for the plugin in .claude-plugin/marketplace.json, so you may want to add that if you'd like the plugin to show up in the Adobe Skills marketplace.

[thagarwal-sjc]

Looks fine, but I don't see an entry for the plugin in .claude-plugin/marketplace.json, so you may want to add that if you'd like the plugin to show up in the Adobe Skills marketplace.

Added the plugin to `.claude-plugin/marketplace.json`