Add SafeSkill security badge (50/100 — Use with Caution)#2
Open
OyaAIProd wants to merge 1 commit into
Open
Conversation
|
@OyaAIProd is attempting to deploy a commit to the KB's projects Team on Vercel. A member of the Team first needs to authorize it. |
laundromatic
added a commit
that referenced
this pull request
Apr 16, 2026
…tioning Replace Core positioning section with Phase 4 locked branding (5 fields): 1-Liner, elevator pitch, thesis, identity, audience. Add supporting one-liners block. Update PR #1 and PR #2 titles + openings. Rewrite PR #2 (LangChain) to use enrich_product with/without strict_confidence_threshold as the autonomy routing pattern. Drop enrich_product_for_autofill references — that tool doesn't exist yet. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
laundromatic
added a commit
that referenced
this pull request
Jun 16, 2026
Three of four architect moves now live in production: - LAU-335 Move 1: cross-tier agreement priced into confidence - LAU-336 Move 2: deleted FIELD_CONFIDENCE_MODIFIERS, replaced with f(method, agreement, has_signal); description-quality heuristics become metadata flags in _shopgraph.quality_signals - LAU-337 Move 3: metric switched from Pearson R to ECE + AUC-ROC Per-field modifier tuning explicitly abandoned (LAU-330 + LAU-333 both regressed and were reverted/canceled). The bar is ECE < 0.10 AND AUC-ROC > 0.75; Pearson R retained as supplementary. Current state (sample 461 post-Move-1): overall AUC 0.626 (was 0.510); 3 of 5 fields above 0.75 AUC bar (brand 0.98, description 0.80, price 0.86). ECE 0.149 will close with LAU-338 Move 4 (isotonic regression once samples accumulate to ~200 per tier x field cell). Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
laundromatic
added a commit
that referenced
this pull request
Jun 18, 2026
CLAUDE.md risk #2: note the GOOGLE_API_KEY 403 was the dominant cause of the degeneracy (now fixed); re-measure before hand-labeling; labeling = B-prime (LAU-353). Session notes appended. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
laundromatic
added a commit
that referenced
this pull request
Jun 18, 2026
CLAUDE.md risk #2 + SESSION_NOTES: ~79% of the corpus ground-truth labels are the extractor's own output (reformatted, commit 267ad75) silently treated as human-verified because provenance was never tracked. KB resetting labels with mandatory per-label origin (human/schema.org/llm/override). Post-fix re-measure is diagnostic-only. Full state in docs/labeling-reset-handoff.md. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
laundromatic
added a commit
that referenced
this pull request
Jun 20, 2026
Representative-proxy correction rate on the non-sourced corpus subset (vs error-skewed sourced contrast). Non-sourced: name 89% / brand 81% / description 76% correct (price/availability freshness-confounded). Mediocre to publish; feeds experiment #2 (coverage vs free incumbents on the un-fed long tail) — the now-decisive gate. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
laundromatic
added a commit
that referenced
this pull request
Jun 26, 2026
Two independent review rounds (coherence + adversarial-after) + KB feedback: - §10 user journeys: fully rewritten to v2.1 (solo-dev first-paste/WS1, BYO, catch-wrong-fact, WS3 exception review); old operator-queue journeys marked history. Honest: freshness = WS1 work-to-build, SOURCE = method attribution not a span/guarantee. - §8 customers: compressed four full vendor write-ups -> tight lighthouse- evidence block (all verified citations + scale numbers preserved); dropped the vestigial "why they'd use / what they get" framing that was off-thesis after the evidence-not-buyer recast. - §6 substrate: ECE+AUC + isotonic tagged internal-only per Option B. - §13 roadmap: benchmark stated carefully (not-worse/inconclusive; formal parity only with hint advantage leveled) instead of flat "at parity". - §14 risk #2: stale AUC figures flagged pre-fix/void; banner scope corrected ("one point" -> the external-gate framing). - Addendum header: "moat" -> "differentiator-work" (4th-sense dedupe). 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
1 participant
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.
🟠 SafeSkill Security Scan Results
Top Findings
src/html-cleaner.ts:47)src/html-cleaner.ts:47)src/html-cleaner.ts:56)src/html-cleaner.ts:56)src/html-cleaner.ts:64)View full report on SafeSkill
About SafeSkill
SafeSkill is a free, open-source security scanner for AI tools, MCP servers, and Claude Code skills. We scan for code exploits, prompt injection, and data exfiltration risks.
False positive? We take accuracy seriously. If any finding above is incorrect, please open an issue and we will fix it immediately.