Bỏ qua tới nội dung
TakoAPI
Quay lại kỹ năng

clawtributor

bởi davida-ps

Security & Passwords
1.5kLượt tải do clawskills.sh công bố
1Sao
0Lượt xem
0 lượt thíchĐăng nhập để thích kỹ năng này
—Chưa có đánh giá
Đánh giá của bạn

Tóm tắt

Community incident reporting for AI agents.

Mô tả

Kỹ năng này làm gì

Clawtributor lets AI agents report security incidents to a community advisory feed. Reports cover malicious prompt attempts, vulnerable skills, and tampering events. All submissions are opt-in and require explicit user approval before being filed as GitHub Issues.

Standardizes security incident reporting across AI agents with a structured review pipeline that publishes numbered advisories (CLAW-YYYY-NNNN), rather than leaving incident documentation ad hoc or siloed per user.

Khi nào nên dùng

  • Reporting a prompt injection attempt observed during an agent session
  • Flagging a skill that sends conversation data to an unknown external server
  • Documenting a tampering attempt against installed security tools
  • Contributing to the shared advisory feed after a suspicious interaction
  • Tracking submitted reports and monitoring their advisory approval status

Quy trình ví dụ

ĐẦU VÀO

User asks: 'I just saw a prompt trying to extract my API keys, can you report it?'

AGENT
  1. 1

    Classify the incident as a malicious prompt attempt and assign severity

  2. 2

    Build a structured JSON report with sanitized evidence and observed indicators

  3. 3

    Present the full report summary to the user and request explicit approval

  4. 4

    Submit the approved report as a GitHub Issue to prompt-security/ClawSec using the security incident template

  5. 5

    Save submission metadata to local state file and return the issue URL to the user

ĐẦU RA

Report filed as GitHub Issue #87, pending maintainer review; if approved, published as advisory CLAW-2026-0087

Yêu cầu

  • GitHub account
  • GitHub CLI (gh) installed and authenticated

Cài đặt

clawhub install davida-ps/clawtributor

Chạy lệnh CLI, hoặc dán prompt vào Claude Code, Cursor, Codex hay coding agent bạn đang dùng.

Liên kết