Skip to content

Review edits: Modern Agents Have Skills Now (for @Roelzz to approve)#332

Merged
adilei merged 9 commits into
mainfrom
post/modern-mcs-agent-skills
Jun 22, 2026
Merged

Review edits: Modern Agents Have Skills Now (for @Roelzz to approve)#332
adilei merged 9 commits into
mainfrom
post/modern-mcs-agent-skills

Conversation

@adilei

@adilei adilei commented Jun 20, 2026

Copy link
Copy Markdown
Collaborator

This is an editor's pass on top of @Roelzz's PR #329, opened as a branch on the main repo so Roel can review and approve the changes before we publish.

Roel: if you're happy with these edits, approve and we'll merge this. If you'd rather keep iterating on your own branch (#329), that's fine too — just let me know and I'll port anything useful over.

What changed

Narrative / framing

  • Reframed the intro around common cases vs. organization-specific procedures at the LLM level (rather than "recalling facts").
  • Dropped the standalone Speed benefit (it double-counted context-management and accuracy); folded a brief speed note into the Accuracy bullet instead.
  • Tightened the context-model diagram caption and cut the dangling "I come back to below" forward-reference.
  • Reworked "A Skill, or a new agent?" into a clear two-signal heuristic: build a separate agent when the capability would stand on its own, or when one agent's tool count starts dragging it down.
  • Reframed "Instructions, or a Skill?" so the "can the agent infer it?" gate is a precondition, and the real decision (always-on vs. situational) leads.
  • Connected the closing of that section back to the earlier benefits (context management, manageability, accuracy/speed).

Technical accuracy

  • Clarified Skill upload options: a standalone SKILL.md, or a .zip bundling resources/scripts (verified against the product).
  • Timestamped per-agent distribution ("as of June 2026") and noted a product Skill catalog is being explored.
  • Explained why tool pointers are "soft" (reference, not a binding or grant).
  • Linked the modern Copilot Studio announcement (no dedicated Skills docs exist yet).
  • Softened the "accuracy/speed degrade as toolset grows" claim to "some data suggests… evaluate for your own agent."

Metadata / reader experience

  • Added agent_edition: modern front matter (required by the build validator; drives the Modern pill) and merged main to bring in the pill rendering.
  • Cut the redundant "shapes" radial Mermaid diagram (the table already covers it).
  • Removed the stretchy mcp tag.
  • Added a closing engagement question.
  • Made the header image alt text whimsical to match the illustration.

Word count ~2,700. All three internal post_url links resolve. Renders locally at HTTP 200.

Roelzz and others added 7 commits June 15, 2026 20:47
…tudioNow

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
…dback

- Replace 'procedural memory' framing with 'instructions and resources loaded on demand'
- Add mental-model analogy section (table + hub diagram) and colored context-flow diagrams
- Update for shipped support: bundled resources and executable scripts
- Address review feedback (intro, accuracy/speed bullets, tool soft-pointing, agent-vs-skill examples, trust, dead link)
- New header image

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
…sections

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
…adata

- Reframe intro around common cases vs. org-specific procedures (LLM-level)
- Drop the standalone Speed benefit; fold a speed note into Accuracy
- Tighten the context-model caption and cut the dangling forward-reference
- Clarify Skill upload options (SKILL.md or .zip with resources/scripts)
- Timestamp per-agent distribution; explain why tool pointers are 'soft'
- Rework 'A Skill, or a new agent?' into a clear two-signal heuristic
- Link the modern Copilot Studio announcement (no Skills docs yet)
- Soften the toolset/accuracy claim and tell readers to evaluate it
- Cut the redundant 'shapes' radial diagram (keep the table)
- Remove the stretchy mcp tag; add a closing engagement question

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
@github-actions

github-actions Bot commented Jun 20, 2026

Copy link
Copy Markdown

Blog preview check passed for 0aa235d.

The Jekyll site compiled successfully and generated previews for the newly added post.

  • \_posts/2026\-06\-15\-modern\-mcs\-agent\-skills\.md -> /mcscatblog/posts/modern\-mcs\-agent\-skills/

Preview artifacts:

Open the workflow run

… heuristic recap

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Restore Speed as a distinct benefit, reframed around Copilot credits

- Bring back Speed as its own benefit bullet (was folded into Accuracy)
- Reframe the credit angle correctly: a Skill nudges the agent toward the
  right approach, reducing knowledge searches, tool calls, and reasoning
  loops, which lowers response time and Copilot credit spend
- Split the closing recap's merged 'Accuracy and speed' into separate
  Accuracy and Speed-and-credits bullets to match the four-benefit framing

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* Tighten opening flow: generic Speed and cost, fold diagram caption

- Rename first benefit bullet to 'Speed and cost' and keep it product-
  generic (latency, throughput, cost); credits stay in the later recap
- Drop the long diagram caption; replace with a short paragraph that
  states the on-demand context model and ties back to the earlier
  benefits, delivering on the section heading

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

---------

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
@adilei adilei merged commit 363eb5a into main Jun 22, 2026
2 checks passed
@adilei adilei deleted the post/modern-mcs-agent-skills branch June 22, 2026 15:45
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants