github sickn33/agentic-awesome-skills v19.13.0
v19.13.0 "Thirteen Reviewed Skills And Security Hardening"

5 hours ago

[19.13.0] - 2026-10-21 - "Thirteen Reviewed Skills And Security Hardening"

Adds 13 catalog skills and ships 2,737 skills. All nine PRs in the maintenance batch were repaired, reviewed and merged.

Added

  • neverapply — find jobs, tailor resumes and prepare applications through the hosted NeverApply MCP service, with user approval before applying.
  • agy-powerline — configure Antigravity CLI's statusline for quotas, model, context and Git information, using a reviewed runtime.
  • calibrate-first — establish a dataset's detection floor and pre-register trading-signal tests; insufficient evidence stays undecidable.
  • lanshu-create-ai-presenter-video — plan presenter-led explainer videos with an immutable upstream source reference and explicit provider, upload and spending consent. The catalog bundles the guide and license, not an executable runtime.
  • apify-youtube-transcript-scraper — retrieve captions, timestamped quotations, SRT/VTT and translated transcripts from YouTube videos, channels and playlists through Apify, with private exports and bounded paid runs.
  • party-planning — estimate food and drink purchases, choose games and draft hosting timelines from disclosed HostPartyPlay heuristics, with allergy, age and privacy limits.
  • pixeltable — build multimodal applications, embedding search and computed-column pipelines with Pixeltable, reviewing application code before schema commands import it.
  • Six official Beatra pointers — ai-audiobook-narration, ai-image-realism, ai-storyboard-generator, wechat-cover-maker, zhongcao-cover-maker and zhongcao-note-copywriter. Each names immutable upstream content and requires complete source review plus consent for credentials, uploads and paid work.

Changed

  • goal-instruction-writer — refresh goal contracts, drift controls and examples to the author's 0.2.5 guidance.
  • datacircle — refresh the published plan-pricing dataset workflow and its usage guidance.
  • apify-tiktok-scraper — require explicit approval of recurring frequency, duration, budget and delivery destination before creating scheduled tracking.
  • Restore 216 skill introduction dates from their actual first-addition history; a previous bulk edit had overwritten them with one date.

Security and reliability

  • Resolve the supplied 93-row security report against current canonical sources: 91 rows corrected, one already corrected and one referencing a retired path. Duplicate mirror findings share their canonical fix; the per-row evidence is recorded in docs/maintainers/security-findings-19.13.0.json.
  • Keep contributor Markdown inert in repository tests: reviewed executable fixtures are maintained separately and document examples are compared as data.
  • Harden local preview, citation, session and state paths against symlinks and traversal; protect Electron driver state with private ownership/permissions, including Windows ACL checks, and atomic writes.
  • Bound template, color and archive parsing; redact sensitive fact-check and PII output; escape terminal controls and neutralize spreadsheet formulas. Tighten credential handling, recurring-work consent, remote-code guidance and external installer prerequisites throughout the affected skills.
  • Patch MCP dependencies in the web application and local plugin server, Vite in the bundled frontend example and redirect handling in the WhatsApp example. Replace the unmaintained ts-node-dev watcher in the WhatsApp and Telegram templates with pinned tsx, removing its vulnerable globbing chain. Audit every canonical npm lockfile. Extend regression coverage for the security boundaries and ensure nested skill trees are included in maintainer review inventories.
  • Verify packed manifests with the bounded runtime archive parser, avoiding the native Windows tar failure; exercise that same check in platform previews.
  • Preserve direct-installer release identity by verifying the actual packed manifest and published registry gitHead; reject staged candidate changes and skipped protected-source validation.

Contributors

Thanks to @msohaildev, @wassim-azi, @assaylabdesk-sudo, @djmahe4, @harshmaur, @tarunbatta, @pierrebrunelle, @beatra-ai, @wu736139669 and @waynehamadi for the catalog contributions and updates.

Don't miss a new agentic-awesome-skills release

NewReleases is sending notifications on new releases.