Repository navigation
docs: add the accelerated evolution skills post and retitle the bias post - #478
Conversation
|
Warning Review limit reached
Next review available in: 19 minutes You've used all free OSS reviews for now. Wait for the free limit to reset to keep reviewing this public repository. How can I continue?After more reviews become available, a review can be triggered using the To avoid repeated limits, reduce automatic review volume by pausing incremental auto-reviews earlier, using label-based review opt-in, excluding WIP or generated PR titles, or requesting reviews manually when the PR is ready. If your team needs uninterrupted high-volume reviews, an organization admin can enable usage-based reviews. How do review limits work?CodeRabbit enforces per-developer PR review limits for each organization. Most developers receive the normal plan review availability. For paid Pro and Pro+ PR reviews, CodeRabbit uses adaptive limits for sustained high-volume activity. When a developer's recent PR review activity reaches the 95th percentile or higher among CodeRabbit users, additional reviews become available more gradually as earlier reviews age out of the rolling window. Please refer docs for additional details. Review details⚙️ Run configurationConfiguration used: Path: .coderabbit.yaml Review profile: CHILL Plan: Pro Plus Run ID: 📒 Files selected for processing (1)
WalkthroughAdded a blog post about agent-skill evaluation and maintenance. Updated a related post’s title, alt text, and sequel link. Added conventions for standalone blog posts and descriptive repository links. ChangesAI skills documentation
Estimated code review effort: 1 (Trivial) | ~5 minutes Possibly related issues
Possibly related PRs
Suggested labels: Suggested reviewers: 🚥 Pre-merge checks | ✅ 5✅ Passed checks (5 passed)
✨ Finishing Touches🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
Preview DeploymentPreview URL: https://f9139633.bestax.pages.dev |
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 1
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@docs/blog/2026-08-05-ai-loop-keeps-skills-honest.md`:
- Line 79: Update the prompt literal in the example to replace both em dashes
with blog-approved punctuation, preserving the prompt’s meaning and wording.
🪄 Autofix
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Pro Plus
Run ID: b646f7a7-f24b-4071-b4a2-9fa2b831f5e7
⛔ Files ignored due to path filters (15)
docs/static/img/ai-loop-keeps-skills-honest-captured-runs.pngis excluded by!**/*.pngdocs/static/img/ai-loop-keeps-skills-honest-captured-runs.svgis excluded by!**/*.svgdocs/static/img/ai-loop-keeps-skills-honest-findings.pngis excluded by!**/*.pngdocs/static/img/ai-loop-keeps-skills-honest-findings.svgis excluded by!**/*.svgdocs/static/img/ai-loop-keeps-skills-honest-gates.pngis excluded by!**/*.pngdocs/static/img/ai-loop-keeps-skills-honest-gates.svgis excluded by!**/*.svgdocs/static/img/ai-loop-keeps-skills-honest-loop.pngis excluded by!**/*.pngdocs/static/img/ai-loop-keeps-skills-honest-loop.svgis excluded by!**/*.svgdocs/static/img/ai-loop-keeps-skills-honest-proof.pngis excluded by!**/*.pngdocs/static/img/ai-loop-keeps-skills-honest-proof.svgis excluded by!**/*.svgdocs/static/img/ai-loop-keeps-skills-honest-shipped.pngis excluded by!**/*.pngdocs/static/img/ai-loop-keeps-skills-honest-shipped.svgis excluded by!**/*.svgdocs/static/img/ai-loop-keeps-skills-honest.pngis excluded by!**/*.pngdocs/static/img/ai-loop-keeps-skills-honest.svgis excluded by!**/*.svgdocs/static/img/fighting-ai-training-bias.svgis excluded by!**/*.svg
📒 Files selected for processing (2)
docs/blog/2026-08-05-ai-loop-keeps-skills-honest.mddocs/blog/2026-08-05-fighting-ai-training-bias.md
![claude[bot]](https://avatars.githubusercontent.com/in/1236702?s=60&v=4){kind=small}
There was a problem hiding this comment.
Deep review — 1 blocking · 1 advisory
| # | Severity | Area | Finding | Location |
|---|---|---|---|---|
| 1 | 🟡 Minor | Correctness | Custom-CSS "10 to 21" excludes i02 (=56); actual i02–i10 range is 10 to 56 | docs/blog/2026-08-05-ai-loop-keeps-skills-honest.md:54 |
| 2 | 🔵 Advisory | Robustness | Em dashes in the ExampleMeta prompt quote technically fall under the blog's no-em-dash-in-demo-strings rule |
docs/blog/2026-08-05-ai-loop-keeps-skills-honest.md:79 |
Overall: Solid, well-sourced docs PR. I verified the headline numbers against eval/skill-loop/report.md (baseline 85 → mean 95.2, median 96, min 89, max 99; -43% cost, -38% turns; 42 raw classNames → 0; 3 grader errors), the rubric's 50-point do-nothing gate against rubric.md, "87 documented components" against the catalog, the "Skills sync (same PR, always)" heading in CONTRIBUTING-COMPONENTS.md, and the skills/CLAUDE.md quotes — all check out. Every issue/PR reference (#367-371, #194-197, #302/#303/#326/#329, #380) resolves to a title matching its description. All 7 SVG+PNG hero pairs exist at 1200×630, each SVG's aria-label mirrors its markdown alt text verbatim, the banner embeds the SVG and section images embed PNGs per the cover contract, and the syndication contract holds (flat .md, /img/-rooted images, no admonitions, the only JSX is inside a tsx fence). The one real defect is a mislabeled range in the results table; the human should focus there.
Residual risk:
- Build/link validation — I could not run
pnpm build/format:checkin this sandbox (pnpm/npx blocked), but I manually confirmed every internal link target exists (/docs/skills/intro,/docs/guides/getting-started/ai-development, both/blog/slugs), soonBrokenLinks: 'throw'should pass; CI still runs the real build. - Other tabular numbers — spot-checked the full report row-by-row; every other cell (baseline, mean, median/min/max, cost, turns, classNames) matches the source. Only the custom-CSS range was off.
- Date-vs-filename convention — the
2026-08-05-prefix matches today; if merge slips past that date the file needs a rename perblog/CLAUDE.md, handled at merge time.
🏄 Clean set, dude — the whole thing rides on receipts and almost all of 'em check out. Just one number wiped out on the CSS wave (56 snuck into a 10-to-21 lineup). Patch that cell and it's all-time, go ahead and paddle it in.
allxsmith
changed the title
|
Rewrote the post top to bottom as a self-contained story ("Accelerated Evolution for Agent Skills") — no series numbering, no internal file/check names as narrative glue, project introduced in a clause, and the bias argument retold in two sentences for readers landing cold from dev.to or Medium. Both review findings are addressed in the rewrite: the custom-CSS row now matches the committed run metrics (56 in the first revised run, then 10 to 21), and the Also in this push: the post file's date bump to 2026-08-06, a regenerated cover headline (ACCELERATED / EVOLUTION; scene unchanged), a reworded sequel line in the bias post, and a new blog voice rule in |
Preview DeploymentPreview URL: https://605893bd.bestax.pages.dev |
There was a problem hiding this comment.
Actionable comments posted: 2
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@docs/blog/2026-08-06-ai-loop-keeps-skills-honest.md`:
- Line 97: Rewrite the closing sentence in the paragraph so it clearly states
that running the evolution loop quickly keeps improving the skills, using the
requested wording and preserving the surrounding content.
- Line 65: Update the sentence in the paragraph beginning “And that's what
‘accelerated’ means here” to replace the ungrammatical training-data time
reference with clear wording that it reaches agents only in a future model.
Preserve the surrounding meaning and proofread the paragraph for clean
mechanics.
🪄 Autofix
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Pro Plus
Run ID: 765ae25d-9961-4da6-8cd9-f81a475257e0
⛔ Files ignored due to path filters (2)
docs/static/img/ai-loop-keeps-skills-honest.pngis excluded by!**/*.pngdocs/static/img/ai-loop-keeps-skills-honest.svgis excluded by!**/*.svg
📒 Files selected for processing (3)
docs/blog/2026-08-05-fighting-ai-training-bias.mddocs/blog/2026-08-06-ai-loop-keeps-skills-honest.mddocs/blog/CLAUDE.md
🚧 Files skipped from review as they are similar to previous changes (1)
- docs/blog/2026-08-05-fighting-ai-training-bias.md
Preview DeploymentPreview URL: https://fc76e517.bestax.pages.dev |
|
@coderabbitai review |
|
There was a problem hiding this comment.
Deep review — 0 blocking · 1 advisory
| # | Severity | Area | Finding | Location |
|---|---|---|---|---|
| 1 | 🔵 Advisory | Correctness | "One npx skills add command drops them into any existing project" reads as installing all seven at once, but the shipped path is one npx skills add … --skill <name> invocation per skill (skills/README.md:27-33); the single-command install is npm create bestax, which the sentence already covers separately. |
docs/blog/2026-08-06-ai-loop-keeps-skills-honest.md:27 |
Overall: This is a sound, carefully-sourced docs PR (one new blog post, a retitle of the prior post, and 7 SVG+PNG image pairs). I chased every quantitative and reference claim to its source and they hold: the score table (85 -> mean 95.2, median 96 / min 89 / max 99; raw CSS 42->0; custom CSS 77->56->range 10-21; $10.55/127 -> mean $6.00/79) matches eval/skill-loop/report.md exactly, the "+/-6 point single-run swing on identical tooling" limit matches report.md:227 and README.md:60, and all twelve issue/PR references (#367-#371, #380, #382, #384, #196, #302/#303/#329) resolve and describe what the prose says. Nothing structural is at risk, just the one prose generalization about the install command above.
Residual risk: the failure class here is inaccurate or broken published content.
- Broken links - refuted: every internal target (
/docs/skills/intro,/docs/guides/getting-started/ai-development,/blog/fighting-ai-training-bias, all/img/*) resolves to a committed source file, andonBrokenLinks: 'throw'gates the build; external links are GitHub/Storybook absolute URLs. - Accessibility regressions - refuted: all 7 new SVG
aria-labels mirror their markdown alt verbatim, and the retitled bias SVG'saria-labelwas updated in lockstep with its alt/title. - Syndication breakage - refuted: the
publish_to_devto: truepost is clean plain markdown (no JSX, no:::admonitions), ships PNG heroes plus an SVG banner per convention, carries the<!-- truncate -->fold, and prettier-check passes.
🏄 Total cruiser of a PR, dude - every number in the set actually catches the wave it claims, the alt-text and aria are riding in perfect sync, and the only ripple is one sentence that makes
skills addsound like a one-and-done. Clean it or don't, it is still good to paddle out.
Preview DeploymentPreview URL: https://3b77a0eb.bestax.pages.dev |
|
🎉 This PR is included in version 5.8.1 🎉 The release is available on: Your semantic-release bot 📦🚀 |
|
🎉 This PR is included in version 4.0.2 🎉 The release is available on: Your semantic-release bot 📦🚀 |
|
🎉 This PR is included in version 1.0.0 🎉 The release is available on: Your semantic-release bot 📦🚀 |
|
🎉 This PR is included in version 2.0.1 🎉 The release is available on: Your semantic-release bot 📦🚀 |
Summary
/blog/ai-loop-keeps-skills-honest), written as one self-contained story: the training-bias problem retold for a stranger, the seven shipped skills, graded cold-start exam runs, ten generations of accelerated evolution (85/100 baseline → 95.2 mean while cost fell 43%), replayable agent runs in Storybook, the review + CI guards that hold the line between generations, and the loop's still-open findings ([Bug] bulma-ui: color props typecheck values with no shipped CSS; Box color falls through to has-text-* #367–[Feature] create-bestax: scaffold polish — gitignore *.tsbuildinfo, PM-neutral next-steps, strictPort fallback #371) plus its stated limits.docs/scripts/pixel-cover-lib.mjsand rasterized throughrasterize:cover; the cover headline reads ACCELERATED / EVOLUTION and every SVGaria-labelmirrors its markdown alt verbatim.aria-label; the drawn headline already matched) and points its sequel line at the new post. Slug and canonical URL unchanged.docs/blog/CLAUDE.md: every post stands alone for a stranger — repo artifacts are link receipts behind descriptive words, never narrative glue.Review findings addressed
eval/skill-loop/runs/*/metrics.json: "56 in the first, then 10 to 21, settling near the ~10-line pattern the skills sanction."ExampleMetaexcerpt: the code fence is gone entirely — the captured-runs section is prose now, which also fit the story-first rewrite better than doctoring a verbatim quote would have.Where this deviates from issue #380
eval/skill-loopon 2026-07-27. The post tells the current story, with every number and quote sourced from the committed harness docs, runs record, and the cited issues/PRs.Verification
pnpm format:check,pnpm lint, and the docs build (onBrokenLinks: 'throw'link validation) all pass.build/.devto-publish/2026-08-06-ai-loop-keeps-skills-honest.mdregenerated with production-URL rewrites and no code fences; dev.to/Medium publishing stays manual post-merge.Closes #380
Summary by CodeRabbit
New Features
Documentation