feat(skill): optimize code exec prompt and add e2e benchmark

[fang-tech]

AgentScope-Java Version

1.0.11-SNAPSHOT

Description

fix #1050, fix #1039, fix #952

• Code execution prompt injection: AgentSkillPromptProvider now appends a
  ## Code Execution section to the skill system prompt when code execution is enabled,
  instructing the LLM to use execute_shell_command with absolute script paths rather than
  describing or rewriting commands inline. The prompt is wired automatically via SkillBox
  on enable() and after uploadSkillFiles().

  • codeExecutionInstruction() builder option: SkillBox.CodeExecutionBuilder exposes a
    new .codeExecutionInstruction(template) method for customizing the injected prompt section.
    %s placeholders are substituted with the uploadDir absolute path (up to 5 times).

  • Remove deprecated SkillBox() no-arg constructor and JarSkillRepositoryAdapter class
    (superseded by ClasspathSkillRepository).

  • E2E benchmark suite (SkillE2ETest): 5 parameterized tests across 4 recall dimensions
    using purpose-built benchmark skills. Tests record pass/fail without individual assertions;
    @AfterAll asserts overall recall rate ≥ 75% and prints a summary table. Benchmark skills:

    • data-transform / data-report — semantic discrimination
    • image-resize — distractor resistance
    • log-parser — fuzzy trigger
    • git-changelog — code execution recall

  • Docs (EN + ZH): add configuration reference for code execution builder options and a new
    "Custom Skill Prompts" section documenting instruction, template, and
    codeExecutionInstruction.

Checklist

Please check the following items before code is ready to be reviewed.

• Code has been formatted with mvn spotless:apply
• All tests are passing (mvn test)
• Javadoc comments are complete and follow project conventions
• Related documentation has been updated (e.g. links, examples, etc.)
• Code is ready for review

[Copilot]

Pull request overview

This PR enhances the Skill system prompt to better trigger actual code execution (via execute_shell_command with absolute paths), adds an E2E benchmark suite for skill recall/code-exec recall, removes deprecated skill-loading utilities, and updates EN/ZH docs accordingly.

Changes:

• Append an optional “Code Execution” section to the skill system prompt when code execution is enabled, with a new codeExecutionInstruction(...) customization hook.
• Add SkillE2ETest benchmark suite + benchmark skills under src/test/resources/e2e-skills/.
• Remove deprecated SkillBox() no-arg constructor and deprecated JarSkillRepositoryAdapter; update documentation.

Reviewed changes

Copilot reviewed 20 out of 20 changed files in this pull request and generated 9 comments.

Show a summary per file

File	Description
docs/zh/task/agent-skill.md	Updates ZH documentation for code execution builder options and custom skill prompts.
docs/en/task/agent-skill.md	Updates EN documentation for code execution builder options and custom skill prompts.
docs/_toc.yml	Adjusts docs navigation entry for the ZH “agent-skill” page.
agentscope-core/src/test/resources/e2e-skills/log-parser/scripts/extract_errors.py	Adds benchmark script for log parsing/code execution recall scenarios.
agentscope-core/src/test/resources/e2e-skills/log-parser/SKILL.md	Adds benchmark skill metadata/doc for log-parser.
agentscope-core/src/test/resources/e2e-skills/image-resize/scripts/resize.py	Adds benchmark script for distractor resistance scenarios.
agentscope-core/src/test/resources/e2e-skills/image-resize/SKILL.md	Adds benchmark skill metadata/doc for image-resize.
agentscope-core/src/test/resources/e2e-skills/git-changelog/scripts/generate_changelog.py	Adds benchmark script for code execution recall scenarios.
agentscope-core/src/test/resources/e2e-skills/git-changelog/SKILL.md	Adds benchmark skill metadata/doc for git-changelog.
agentscope-core/src/test/resources/e2e-skills/data-transform/scripts/json_to_csv.py	Adds benchmark script for semantic discrimination scenarios.
agentscope-core/src/test/resources/e2e-skills/data-transform/scripts/csv_to_json.py	Adds benchmark script for semantic discrimination scenarios.
agentscope-core/src/test/resources/e2e-skills/data-transform/SKILL.md	Adds benchmark skill metadata/doc for data-transform.
agentscope-core/src/test/resources/e2e-skills/data-report/scripts/summarize.py	Adds benchmark script for semantic discrimination scenarios.
agentscope-core/src/test/resources/e2e-skills/data-report/SKILL.md	Adds benchmark skill metadata/doc for data-report.
agentscope-core/src/test/java/io/agentscope/core/skill/SkillBoxTest.java	Adds integration tests ensuring prompt injection behavior when code execution is enabled.
agentscope-core/src/test/java/io/agentscope/core/skill/AgentSkillPromptProviderTest.java	Adds unit tests for code execution prompt section behavior and customization.
agentscope-core/src/test/java/io/agentscope/core/e2e/SkillE2ETest.java	Adds concurrent-gated E2E benchmark tests for skill recall + code execution recall.
agentscope-core/src/main/java/io/agentscope/core/skill/util/JarSkillRepositoryAdapter.java	Removes deprecated adapter (superseded by ClasspathSkillRepository).
agentscope-core/src/main/java/io/agentscope/core/skill/SkillBox.java	Removes deprecated no-arg constructor; wires code-exec prompt settings from builder.
agentscope-core/src/main/java/io/agentscope/core/skill/AgentSkillPromptProvider.java	Implements conditional code execution section injection + customization plumbing.

[codecov]

Codecov Report

❌ Patch coverage is 85.71429% with 3 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
...gentscope/core/skill/AgentSkillPromptProvider.java	85.71%	0 Missing and 2 partials ⚠️
...c/main/java/io/agentscope/core/skill/SkillBox.java	85.71%	0 Missing and 1 partial ⚠️

📢 Thoughts on this report? Let us know!