Repository navigation
Conversation
Sample 1: capital expenditures over the three most recent fiscal quarters reported as of 2026-11-06 (Amazon, Alphabet, Microsoft; 10-Q/10-K cash-flow statements). Sample 2: launch order of the 2026 always-on personal agents (Gemini Spark, Grok Bot, Muse; company launch posts). Generated with [Claude Code](https://claude.ai/code) via [HappyHerd](https://github.com/NickGuAI/HappyHerd) Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Each team evaluates its Studio 06 deep research system in Harbor: 10 research questions with a known answer and a pass rule, five trials per question, one LLM judge call per trial (Anthropic's clear-answer method), pass@1 and pass^5. Four checkpoints of two 1-point lines and a 2-point explanation, 10 points. Due Friday, November 13, 2026, 13:10 ET. Generated with [Claude Code](https://claude.ai/code) via [HappyHerd](https://github.com/NickGuAI/HappyHerd) Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Generated with [Claude Code](https://claude.ai/code) via [HappyHerd](https://github.com/NickGuAI/HappyHerd) Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Generated with [Claude Code](https://claude.ai/code) via [HappyHerd](https://github.com/NickGuAI/HappyHerd) Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
- Define Harbor, trial, installed agent, oracle, wrong-answer agent, reward, pass@1 and pass^5 where they first appear. - Copy the submission template first, then the kit files by name. Copy agent_template.py to agent.py and run it with -a agent:TeamAgent. - Add the copy commands for eval/oracle/pass, eval/oracle/fail and eval/job, and examples of judge.json and results.json. - Shorten rubric lines 1.1, 2.2 and 4.1. Read 4.1 from eval/job/*/verifier/judge.json. Drop the answer-file clause. - Name the five products in sample q02, call capex a stand-in in q01, and keep only the three answers in each key. - Point the Harbor readings at docs.harborframework.com pages that resolve. Split every sentence that used a semicolon. Generated with [Claude Code](https://claude.ai/code) via [HappyHerd](https://github.com/NickGuAI/HappyHerd) Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
- Rubric 2.2 asks for one known answer per question that its pass rule checks. Rubric 4.1 names the kit's judge model, gpt-5.5. - The kit README asks for the 10 questions before the tasks are built and keeps a sample check in its own folder and job names. Field lists replace the long sentences. - Say where the answers are for the hand labels. Use "Docker sandbox" and "due date" throughout. Shorter sentences in the submission guide. Generated with [Claude Code](https://claude.ai/code) via [HappyHerd](https://github.com/NickGuAI/HappyHerd) Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Each member submits the same team pull-request link through CourseWorks, as in Studio 06, and the private repository is the one bootstrap.sh created, as in Studio 05. Generated with [Claude Code](https://claude.ai/code) via [HappyHerd](https://github.com/NickGuAI/HappyHerd) Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Studio 09: Build Your Eval
Each team evaluates the deep research system it built in Studio 06: 10 research questions with a known answer, five Harbor trials per question, one LLM judge call per trial (Anthropic's clear-answer method), pass@1 and pass^5.
Files
Checkpoints
BaseInstalledAgent· 1.2 the grader check: oracle 1, wrong answer 0eval/job/· 3.2results.jsonmatches the trialsjudge.json: the kit’s judge model (gpt-5.5), rubric hash, five scores, reason, verdict equal to the reward · 4.2 10 hand labels beside the judge's verdicts🤖 Generated with Claude Code