[17.0.0] - 2026-09-08 - "Evidence, Portable Bundles, and Grounded Profiles"
A skill-library release for Claude Code, Cursor, Codex CLI and Gemini CLI,
with bounded repair evidence, conservative local bundle adaptation, and a more
conversational LinkedIn profile workflow. The catalog contains 2,115 skills.
Existing agent-owned selection and installation interfaces remain supported.
Start here
- Install:
npx agentic-awesome-skills@17.0.0 - Choose your tool
- Best skills by tool
- Bundles
- Workflows
Added
break-ai-fix-loops: a three-attempt repair workflow with stable symptom
fingerprints, real-path evidence, negative controls and disposable rollback
checks. Its local Python helper validates fingerprint records.skill-porter: preview conservative tool-name substitutions and copy complete
local skill bundles to fresh destinations. Preserve support bytes and source
metadata; reject links, special files and overwrites. Remote acquisition and
semantic adaptation remain separate reviewed steps, not automatic conversion.
Improved
- LinkedIn profile rewrites collect missing source material conversationally,
honor the user's delivery preferences and ground skills and achievements in
evidence. Supporting research uses relevant public source links; illustrative
metrics are explicitly fictional. - Preserve upstream MIT license notices in both new bundles.
- Permit the two reviewed script-bearing contributions through a protected-base
ledger bound to their complete skill trees. Exact-current-head review, source
checks, mode/path/size restrictions and strict branch protection remain required.
Validation scope
Local repository checks, bundled helper tests and a real temporary-workspace
porter invocation cover the changed behavior. Text translation does not certify
Antigravity tool compatibility or execute imported skills. GitHub security and
root/web dependency audits were rechecked during maintenance.
Thanks to @twoicewoo (#1337), @Pranav-Nexus (#1413), and
@WHOISABHISHEKADHIKARI (#1379) for their contributions.