本文へ移動
cccskills
無料GitHub で公開

swe-agent-issue-loop

Use when turning GitHub issues or bug reports into reproducible coding-agent repair loops with sandboxed execution, focused tests, patch generation, and regression verification.

mouadja02/skills
インストール方法を見る

含まれるファイル（3）

  • SKILL.md3.3 KB
  • references/agent-loop-checklist.md1.3 KB
  • scripts/issue_repro_matrix.py2.0 KB

SKILL.md（原文）

インストールする前に、エージェントに与えられる指示の中身を確認できます。

SWE Agent Issue Loop

Use this skill to handle issue-driven software engineering tasks the way modern coding-agent benchmarks and production agent harnesses do: isolate the repository, reproduce the failure, edit narrowly, and verify with executable tests.

When To Use

Activate when the user asks to:

  • Fix a GitHub issue, bug report, failing test, or CI failure
  • Build a coding-agent benchmark task
  • Convert an issue into a reproducible repair workflow
  • Compare agent scaffolds on software engineering tasks
  • Add guardrails around autonomous coding agents

Repair Loop

  1. Frame the issue

    • Extract expected behavior, observed behavior, affected files, environment, and acceptance criteria.
    • Identify what evidence is missing before editing.
    • Record any user constraints such as no refactor, no dependency changes, or no UI changes.

  2. Build a reproduction

    • Run the smallest relevant failing command.
    • Prefer a focused test over full-suite execution.
    • If no test exists, write or sketch one before implementation when feasible.
    • Store command, result, and failure reason.

  3. Patch narrowly

    • Read the local code around the failing behavior.
    • Change the smallest coherent surface.
    • Avoid solving benchmark tasks by hardcoding issue text, test names, or environment-specific values.

  4. Verify

    • Re-run the reproducer.
    • Run adjacent tests or a targeted suite.
    • Check for formatting, lint, type, or build regressions when relevant.
    • Summarize command outcomes in the final response.

  5. Harden the harness

    • Use sandboxed execution for autonomous agents.
    • Cap command timeouts and output length.
    • Separate model instructions from tool output.
    • Preserve full trajectories for later review.

Benchmark-Aware Guidance

SWE-bench Verified emphasizes clear issue descriptions, correct tests, and solvable tasks. Mirror that standard for internal benchmarks:

  • Do not include ambiguous reports without expected behavior.
  • Do not count a task as solved unless tests or deterministic checks pass.
  • Track whether the agent used only allowed context.
  • Capture the diff and trajectory, not only pass/fail.

Helper Script

Use issue_repro_matrix.py to convert issue text into a reproducibility matrix:

python scripts/issue_repro_matrix.py issue.md

References

Read agent-loop-checklist.md before designing a coding-agent harness or benchmark task.

External grounding:

レビュー

まだレビューはありません。使ってみた感想をお寄せください。

同じリポジトリのスキル

概要と使いどころ

ace-context-evolution

無料

Use when long-lived agents, project memories, system prompts, or playbooks are repeatedly summarized, rewritten, or optimized and risk losing domain detail, provenance, or execution learnings.

日本語の概要は準備中です。原文の説明を表示しています。

mouadja02/skills102026年10月16日 更新

acquire-codebase-knowledge

無料

Use this skill when the user explicitly asks to map, document, or onboard into an existing codebase. Trigger for prompts like "map this codebase", "document this architecture", "onboard me to this repo", or "create codebase docs". Do not trigger for routine feature implementation, bug fixes, or narrow code edits unless the user asks for repository-level discovery.

日本語の概要は準備中です。原文の説明を表示しています。

mouadja02/skills102026年10月16日 更新

add-educational-comments

無料

Adds educational comments to a code file so it becomes a learning resource, asking for the file (with a list of close matches) if none is given. Use when preparing code for learners, onboarding, or teaching.

日本語の概要は準備中です。原文の説明を表示しています。

mouadja02/skills102026年10月16日 更新

adobe-illustrator-scripting

無料

Write, debug, and optimize Adobe Illustrator automation scripts using ExtendScript (JavaScript/JSX). Use when creating or modifying scripts that manipulate documents, layers, paths, text frames, colors, symbols, artboards, or any Illustrator DOM objects. Covers the complete JavaScript object model, coordinate system, measurement units, export workflows, and scripting best practices.

日本語の概要は準備中です。原文の説明を表示しています。

mouadja02/skills102026年10月16日 更新

agent-governance

無料

Patterns and techniques for adding governance, safety, and trust controls to AI agent systems. Use this skill when: - Building AI agents that call external tools (APIs, databases, file systems) - Implementing policy-based access controls for agent tool usage - Adding semantic intent classification to detect dangerous prompts - Creating trust scoring systems for multi-agent workflows - Building audit trails for agent actions and decisions - Enforcing rate limits, content filters, or tool restrictions on agents - Working with any agent framework (PydanticAI, CrewAI, OpenAI Agents, LangChain, AutoGen)

日本語の概要は準備中です。原文の説明を表示しています。

mouadja02/skills102026年10月16日 更新

agent-owasp-compliance

無料

Check any AI agent codebase against the OWASP Agentic Security Initiative (ASI) Top 10 risks. Use this skill when: - Evaluating an agent system's security posture before production deployment - Running a compliance check against OWASP ASI 2026 standards - Mapping existing security controls to the 10 agentic risks - Generating a compliance report for security review or audit - Comparing agent framework security features against the standard - Any request like "is my agent OWASP compliant?", "check ASI compliance", or "agentic security audit"

日本語の概要は準備中です。原文の説明を表示しています。

mouadja02/skills102026年10月16日 更新

mouadja02 のスキルをすべて見る

このスキルの問題を報告する