Add Bounty Sieve guardrail tool

[junbuilds96]

Resource Addition

Category: Tools and Frameworks
Type: Tool

Resource Details

• Title: Bounty Sieve
• URL: https://github.com/junbuilds96/bounty-sieve
• Authors/Source: junbuilds96
• Year: 2026

Why This Should Be Included

Bounty Sieve is directly related to agent, tool, and skill security because it gives coding agents a read-only intake gate before they act on bounty-like GitHub issues. It helps catch prompt/context exfiltration, credential/wallet requests, star-gated tasks, duplicate-PR traps, and vague high-reward work before the agent clones code, comments, claims work, or opens PRs.

It ships a SKILL.md, explicit human approval gates, offline fixtures, GitHub issue and URL-list import, deterministic triage, and local decision briefs.

Checklist

• Resource is directly related to agent/tool/skill security
• Not already listed in the README
• Follows the formatting guidelines in CONTRIBUTING.md
• Link is accessible and working

[junbuilds96]

Update: Bounty Sieve now has a static synthetic sample report that is easier to scan than the Markdown case study:

https://github.com/junbuilds96/bounty-sieve/blob/main/examples/synthetic-report.html

It shows the bundled fixture outcome at a glance: 2 pursue / 2 watch / 3 reject, including prompt/context exfiltration, wallet/secret access, star-gated reward, and duplicate-PR swarm examples. The data is explicitly labeled synthetic and not live bounty data.