Name: Repository Issue Solver
Brand: Tetrees AI
Availability: InStock

Tetrees Agent guide

What this agent does

Coding assistants often edit before reproducing, change tests to fit a patch, touch unrelated files, and report success without rerunning the exact failure path or reviewing the final diff.

Designed for

Maintainers triaging repository issues
Engineering teams preparing patch plans
Developers reviewing an agent-generated change
QA engineers converting reports into regressions

Inputs it works with

Repository snapshot or relevant source attachments
Issue or pull-request description
Observed error, environment, and reproduction steps
Project-specific test and style commands

Outputs you receive

Issue contract and reproduction plan
Relevant-file map and causal hypothesis
Minimal patch plan or review
Regression command matrix
Final diff risk and closure checklist

How the workflow runs

1Restate the acceptance criteria and identify missing reproduction details.
2Map the smallest relevant code surface before suggesting edits.
3Create or describe a focused reproduction that fails for the stated reason.
4Trace the causal path and propose the smallest non-test change that satisfies the contract.
5Run or specify targeted tests, type checks, lint, and the original reproduction after the change.
6Review the final diff for test tampering, unrelated edits, cleanup artifacts, and uncovered edge cases.

Example usage

Start with one of these requests, then replace the details with your own.

Turn a flaky issue into a reproducible patch plan

A maintainer receives a bug report that appears only with an empty cache after a version upgrade.

Example request

Read the attached issue and source excerpts. Identify the likely state transition, write a minimal reproduction plan, and propose the smallest source-only fix with regression checks.

Expected outcome: Acceptance criteria, a reproducible failing path, causal file map, minimal patch proposal, edge cases, and commands that prove the original failure is fixed.

Audit an agent-generated pull request

A team needs to decide whether a generated patch actually solves the issue without weakening tests.

Example request

Review this issue, diff, and test output. Check for test manipulation, unrelated changes, missing failure reproduction, and regressions. Give a merge or revise verdict.

Expected outcome: A traceable requirement-to-diff review, test-integrity check, risk list, missing evidence, and a bounded merge recommendation.

A successful run should

✓ The original failure is reproduced or explicitly blocked
✓ The proposed change is minimal and causal
✓ Tests are not changed to conceal failure
✓ Relevant checks are rerun after the final edit
✓ The closure report distinguishes observed from assumed

Boundaries and limitations

• Hosted mode cannot edit or execute a repository unless a client-side audited extension supplies that capability
• Static excerpts may omit the real cause
• Generated patches require maintainer review
• Security-sensitive changes need a dedicated security review

能力

issue triagerepository navigationreproduction designpatch planningregression review

Choose OpenAI, Claude or Z.AI in Agent Studio, then run with Tetrees Points or BYOK.

Execution tools

Hosted runs use only this signed set. Local MCP may add tools that you separately configure and approve.

recall pack memorypropose pack growthlist run attachmentsread run attachmentweb search

Agent AVCP

Agent AVCP レポート

Tested v1.0.0

テスト済み TAIP 成果物は本レポートと署名に結合されています。スコアは試験証拠であり将来の全出力を保証しません。

Overall score

9.5

out of 10

Security9.5

Function10.0

Completion10.0

Mandatory gates

provenance✓ Pass

costCeilings✓ Pass

growthRollback✓ Pass

skillExecution✓ Pass

memoryIsolation

実行または取得にはログインしてください。

レビュー

まだレビューがありません。

Repository Issue Solver