How does it manage Claude token costs?

It uses a tiered approach, delegating heavy data parsing to the smaller Haiku model while only using the primary agent for final decision-making and drafting fixes.

What specifically does review-reviewers track?

It evaluates the 'public outcomes' of the Tend bot, checking if its PRs are merged or reverted, and if its reviews are acted upon or corrected by human maintainers.

Does this skill require special GitHub permissions?

Yes, it requires a bot token with 'gist' scope to record evidence and permissions to create issues and PRs on the target repositories.

Where are the analysis findings stored?

Findings are stored in secret GitHub Gists owned by the bot, which are then indexed in a monthly tracking issue on the main Tend repository for maintainer visibility.

Tend CI Outcome Monitor

Name: Tend CI Outcome Monitor
Author: max-sixty

bymax-sixty

•

分析与监控

Monitors and analyzes the real-world impact of Tend CI bot outputs to ensure high-quality, accepted contributions.

The review-reviewers skill provides an hourly, outcome-based audit of the Tend CI bot's behavior within GitHub repositories. By focusing on whether the bot's PRs, reviews, and comments are accepted or rejected by human maintainers, it identifies behavioral patterns without the high token costs of constant log parsing. It utilizes a hierarchical agent approach, delegating broad data gathering to Haiku subagents while reserving higher-order reasoning for identifying systemic issues. When performance gaps are found, the skill automatically tracks evidence in GitHub Gists and can escalate findings into actionable PRs or issues to refine bot logic.

主要功能

01Outcome-based analysis of automated reviews and PR acceptance rates

02Cost-efficient exploration using Haiku subagents for data gathering

0314 GitHub stars

04Persistent evidence tracking via secret GitHub Gists and monthly issues

05Repo-specific guidance prioritization to respect local contribution rules

06Automated escalation of behavioral problems to the core Tend repository

使用场景

01Auditing AI-powered CI workflows for recurring 'hallucinations' or rejected logic

02Tracking the long-term ROI and acceptance rate of automated developer tools

03Automating the quality assurance of AI agents performing code reviews and triage

主要功能

01Outcome-based analysis of automated reviews and PR acceptance rates

02Cost-efficient exploration using Haiku subagents for data gathering

0314 GitHub stars

04Persistent evidence tracking via secret GitHub Gists and monthly issues

05Repo-specific guidance prioritization to respect local contribution rules

06Automated escalation of behavioral problems to the core Tend repository

使用场景

01Auditing AI-powered CI workflows for recurring 'hallucinations' or rejected logic

02Tracking the long-term ROI and acceptance rate of automated developer tools

03Automating the quality assurance of AI agents performing code reviews and triage