docs: more web skills + score->fix->re-score case studies#46
Merged
Conversation
…udies Expanded the from-the-web run to 123 community skills across four public repos (added anthropics/skills). Every skill was safety-scanned as inert text first — 0/123 tripped the gate, none executed. New real before->after case studies (full skill dirs, verbatim, attributed): - daydream (glebis): no frontmatter at all, 38/F/Reject -> 100/A/Ship (+62) - disk-cleanup (glebis): invalid YAML frontmatter, 50/D -> 100/A (+50) - rehydrate (glebis): <> in description, 92/A -> 100/A (+8) Each "after" is the actual re-score of the fixed skill with zero remaining findings. Kept in docs/examples; from-the-web.md now leads with a four-case table. Synced the 69 -> 123 references on the benchmark README and landing lead. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Imports more bad skills from the web, runs them through skval, and adds the improvements to the example pages.
Safety
Expanded run to 123 community skills across 4 public repos. Every one was safety-scanned as inert text first — 0/123 tripped the gate, none executed (only SKILL.md text read; symlink artifacts skipped).
New case studies (real, verbatim, attributed —
docs/examples/skill-benchmark/)daydream(glebis)disk-cleanup(glebis)rehydrate(glebis)<>in descriptionEach was fetched as a full skill directory (so refs resolve — no single-file artifacts), scored verbatim, fixed against skval's ranked finding, and re-scored to a clean 100 (zero remaining findings).
from-the-web.mdnow leads with a four-case table (these + the existingannotate).Kept in
docs/examples(not headlined on the landing page). 69 → 123 references synced.🤖 Generated with Claude Code