feat: improve openspec-apply-change skill score 74% → 90%

[yogesh-tessl]

Hey @benkli01 👋

ML Inference Advisor is a cool concept, helping developers evaluate model compatibility for different inference targets before they're too deep in. The OpenSpec workflow reference in your AGENTS.md shows you're thinking about structured agent collaboration, which is great to see in a hardware-adjacent tool.

ran your skills through tessl skill review at work and found some targeted improvements for openspec-apply-change. Here's the before/after:

Skill	Before	After	Change
openspec-apply-change	74%	90%	+16%

Changes made

Description improvements (biggest impact - 57% → 100%):

• Rewrote the description to list concrete actions: reading task definitions, generating code changes, marking tasks complete, tracking progress
• Added explicit "Use when..." clause with natural trigger terms: "start or resume implementation", "spec tasks", "task list", "next pending change"
• Added an inline example invocation to the input section for discoverability

Content improvements (77% → 77%, maintained while reducing bulk):

• Consolidated three verbose output format blocks (implementation, completion, pause) into a single concise output format paragraph - same information, ~60% fewer lines
• Merged redundant pause/blocker conditions that were duplicated between step 6 and the Guardrails section into one consolidated list
• Tightened the Fluid Workflow Integration section from a bulleted pair into a single focused paragraph
• Kept both .github/skills/ and .codex/skills/ copies in sync

What stayed the same:

• All 7 workflow steps preserved with their CLI commands and state handling logic
• Guardrail intent unchanged - just consolidated
• OpenSpec-specific terminology and domain framing preserved throughout

---

also stress-tested your openspec-apply-change skill against a few real-world task evals and it held up really well on multi-schema task iteration with blocked-state recovery. Kudos for that.

quick honest disclosure. I work at https://github.com/tesslio where we build tooling around skills like these. Not a pitch, just saw room for improvement and wanted to contribute.

if you want to self-improve your skills, or define your own scenarios to pressure test, just ask your agent (Claude Code, Codex, etc.) to evaluate and optimize your skill with Tessl. Ping me @yogesh-tessl, if you hit any snags.