refactor: collapse the three HTML skills into one effective-html router#6
Open
kylesnowschwartz wants to merge 2 commits into
Open
refactor: collapse the three HTML skills into one effective-html router#6kylesnowschwartz wants to merge 2 commits into
kylesnowschwartz wants to merge 2 commits into
Conversation
Merge the html, html-diagram, and html-plan skills into a single effective-html skill that routes between artifact, diagram, and plan modes. The three skills shared an identical reference corpus and the same dark-mode requirement, and each bundled its own copy of references/html-effectiveness/, so this dedups ~24k lines of duplicated examples. - New skills/effective-html/SKILL.md routes by mode (artifact / diagram / plan) - Single references/html-effectiveness/ plus architecture-example.html - Drop disable-model-invocation so the model can reach for it - Update plugin.json, marketplace.json, skills.sh.json, and README
Point the codex agent interface at $effective-html covering all three modes, and enable implicit invocation to match dropping disable-model-invocation on the Claude side. Update the codex manifest description.
Contributor
|
Okay, this is a fair change. I'll review this weekend seems like a fine approach though |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Thanks for this plugin, and for Plannotator, they're really effective.
I've got a suggestion, and I think it applies to your other plugins too: make the skills model-invokable, and at the same time fold very similar skills into one routing-table skill with a single, broader description.
Here's my reasoning.
Right now the three HTML skills are command-only (
disable-model-invocation: true), so I have to know to type/html-diagram. That works, but it puts the routing on me. If the skill is model-invokable I can just say "make me an HTML file that..." and the right mode fires on its own.The catch is that model-invocation and three-separate-skills don't sit well together. While they're command-only the overlap is harmless, I pick the one I want by name. But once the model is choosing, three descriptions that all say "produce an effective HTML artifact" compete for the same job. That costs context budget and lowers activation, the model's picking between near-identical descriptions so it's likelier to land on the wrong mode, or none. Folding them into one skill with a mode table sorts that out: one description gets the skill to fire, then the table picks artifact vs diagram vs plan.
Separately, and regardless of the above, the examples were literally triplicated. Each skill carried its own byte-identical copy of the
html-effectiveness/corpus (I checked withdiff -r), so the repo held three copies of the same ~25 files.What's in the PR:
effective-htmlskill with a routing table (artifact / diagram / plan) over shared setup. Each mode keeps its original guidance, including the diagram's pointer toarchitecture-example.htmland the plan's "keep the writing close to the input" note.references/html-effectiveness/plusarchitecture-example.html, dropping the two duplicate copies (~24k lines). The reference content itself is untouched..claude-plugin/plugin.json,marketplace.json,skills.sh.json, the codex.codex-plugin/plugin.json, andagents/openai.yaml.