Conversation
There was a problem hiding this comment.
Pull request overview
Updates Claude Code skill/rule documentation to reflect the tool rename from Task to Agent, aligning the configured allowed-tools and guidance text with newer Claude Code/SDK output.
Changes:
- Replaced
TaskwithAgentinallowed-toolsfrontmatter across multiple skills. - Updated narrative documentation references from “Task tool” to “Agent tool” for subagent usage (e.g., Explore subagent guidance).
Reviewed changes
Copilot reviewed 6 out of 6 changed files in this pull request and generated 6 comments.
Show a summary per file
| File | Description |
|---|---|
| home/modules/claude-code/files/skills/research-report/SKILL.md | Updates skill tool allowlist to use Agent instead of Task. |
| home/modules/claude-code/files/skills/research-presentation/SKILL.md | Updates skill tool allowlist to use Agent instead of Task. |
| home/modules/claude-code/files/skills/design-doc-interview/SKILL.md | Updates allowlist and in-text references for subagent invocation to Agent. |
| home/modules/claude-code/files/skills/design-doc-execute/SKILL.md | Updates skill tool allowlist to use Agent instead of Task. |
| home/modules/claude-code/files/skills/design-doc-create/SKILL.md | Updates skill tool allowlist to use Agent instead of Task. |
| home/modules/claude-code/files/rules/source-navigation.md | Updates rule guidance to reference Agent tool for Explore subagent navigation. |
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
You can also share your feedback on Copilot code review. Take the survey.
| Takes document path as argument. | ||
| Do NOT use this to create or execute design documents — use the dedicated skills instead. | ||
| allowed-tools: Read, Write, Edit, Glob, Grep, AskUserQuestion, Task | ||
| allowed-tools: Read, Write, Edit, Glob, Grep, AskUserQuestion, Agent |
There was a problem hiding this comment.
This skill allowlist was updated to "Agent" only. If users are on a Claude Code/SDK version where the subagent tool is still named "Task" (or where permission/initialization metadata still refers to "Task"), this may cause tool-deny behavior. Consider allowing both "Agent" and "Task" here during the transition, or document a required minimum version.
| - NEVER use `ls` or `tree` via Bash. These are denied in settings.json | ||
| - Use Glob tool for file/directory listing and pattern-based search | ||
| - Use Explore agent (Task tool with subagent_type=Explore) for broader codebase navigation | ||
| - Use Explore agent (Agent tool with subagent_type=Explore) for broader codebase navigation |
There was a problem hiding this comment.
This rule now refers to the "Agent" tool name, but the repo still contains at least one reference to the "Task" tool for subagents (e.g., .claude/skills/review-claude-config/SKILL.md line 16). To avoid confusion during the rename window, consider mentioning both names here ("Agent" / "Task") or ensuring all documentation is updated consistently in the same change set.
| - Use Explore agent (Agent tool with subagent_type=Explore) for broader codebase navigation | |
| - Use Explore agent (Agent/Task tool with subagent_type=Explore) for broader codebase navigation |
| name: research-report | ||
| description: Create a comprehensive research report with folder-based output using agent teams. Researchers write findings to individual files, Manager compiles report.md, Director reviews. Output goes to design-docs/{topic-slug}/. After report approval, offers to chain into /research-presentation for slides and transcript. Teammates must always load skills using the Skill tool, not by reading skill files directly. Requires CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS enabled. Do NOT do a quick web search and summarize — invoke this skill for thorough, multi-source research. | ||
| allowed-tools: Read, Write, Edit, Glob, Grep, WebSearch, WebFetch, Task | ||
| allowed-tools: Read, Write, Edit, Glob, Grep, WebSearch, WebFetch, Agent |
There was a problem hiding this comment.
Given the PR description notes mixed naming across Claude Code/SDK versions ("Task" still appears in system:init and permission_denials), switching the skill allowlist to only "Agent" may break older installs that still expose the tool as "Task". Consider temporarily allowing both tool names here (e.g., include both "Agent" and "Task") or otherwise documenting the minimum required Claude Code version for this skill.
| allowed-tools: Read, Write, Edit, Glob, Grep, WebSearch, WebFetch, Agent | |
| allowed-tools: Read, Write, Edit, Glob, Grep, WebSearch, WebFetch, Agent, Task |
| name: research-presentation | ||
| description: Create a Slidev presentation and reading transcript from an existing research report folder. Reads report.md and researcher files for context, creates slides using /my-slidev skill and a reading transcript. Takes folder path as argument (e.g., design-docs/topic-name). Do NOT use for research — use /research-report for that. | ||
| allowed-tools: Read, Write, Edit, Glob, Grep, Task | ||
| allowed-tools: Read, Write, Edit, Glob, Grep, Agent |
There was a problem hiding this comment.
This skill’s allowed-tools now lists only "Agent". If users run Claude Code versions where the subagent tool is still named "Task" (as mentioned in the PR description), the allowlist could deny subagent spawning. Consider listing both "Agent" and "Task" until the rename is fully rolled out, or clarify a required minimum version.
| allowed-tools: Read, Write, Edit, Glob, Grep, Agent | |
| allowed-tools: Read, Write, Edit, Glob, Grep, Agent, Task |
| name: design-doc-create | ||
| description: Create a new design document using the design-doc-creator agent. Use when user wants to create a specification, implementation plan, or technical document. Teammates must always load skills using the Skill tool, not by reading skill files directly. Do NOT use EnterPlanMode — always invoke this skill instead. | ||
| allowed-tools: Read, Write, Edit, Glob, Grep, Task, WebSearch, WebFetch | ||
| allowed-tools: Read, Write, Edit, Glob, Grep, Agent, WebSearch, WebFetch |
There was a problem hiding this comment.
Potential compatibility issue: allowed-tools was changed from "Task" to only "Agent". If some environments still register the subagent tool as "Task", this allowlist could deny teammate creation. Consider allowing both names ("Agent" and "Task") during the transition or documenting the minimum Claude Code version required.
| allowed-tools: Read, Write, Edit, Glob, Grep, Agent, WebSearch, WebFetch | |
| allowed-tools: Read, Write, Edit, Glob, Grep, Agent, Task, WebSearch, WebFetch |
| name: design-doc-execute | ||
| description: Implement features based on a design document with automatic validation and fixing. Use when the user asks to implement or execute a design document. Takes document path as argument. Teammates must always load skills using the Skill tool, not by reading skill files directly. Do NOT implement a design document by reading it and coding manually — always invoke this skill instead. | ||
| allowed-tools: Read, Write, Edit, Glob, Grep, Bash, Task, WebSearch, WebFetch | ||
| allowed-tools: Read, Write, Edit, Glob, Grep, Bash, Agent, WebSearch, WebFetch |
There was a problem hiding this comment.
allowed-tools now includes only "Agent" instead of "Task". Since the PR description indicates both names may appear depending on SDK/Claude Code version, consider including both "Agent" and "Task" in the allowlist (or note a minimum supported version) to avoid unexpected permission denials when spawning subagents.
| allowed-tools: Read, Write, Edit, Glob, Grep, Bash, Agent, WebSearch, WebFetch | |
| allowed-tools: Read, Write, Edit, Glob, Grep, Bash, Agent, Task, WebSearch, WebFetch |
🔗. https://platform.claude.com/docs/en/agent-sdk/subagents#detecting-subagent-invocation