feat(agents): add activation profiles for improved agent routing#56
Conversation
There was a problem hiding this comment.
Pull request overview
This PR enhances agent discoverability by adding activation profiles to all 18 agents across three platforms (VSCode, Copilot CLI, and Claude). The changes improve how AI models select the appropriate specialist agent by providing keyword-based activation vocabulary and conversational summoning prompts. Each agent now has an "Activation Profile" section with 20 domain-specific keywords and a persona-driven description that helps models understand when and how to invoke each specialist.
Key Changes
- Updated
descriptionfields in YAML frontmatter with more activation-focused language that emphasizes agent capabilities and use cases - Added "Activation Profile" sections to all agents containing Keywords (20-word vocabulary cluster) and Summon (conversational invocation prompt)
- Maintained consistency across all 54 agent files (18 agents × 3 platforms) through template-based regeneration
Reviewed changes
Copilot reviewed 72 out of 72 changed files in this pull request and generated 1 comment.
| File | Description |
|---|---|
templates/agents/*.shared.md (18 files) |
Source templates with new activation profiles that propagate to all platforms |
src/vs-code-agents/*.agent.md (18 files) |
VSCode agent definitions regenerated from templates with platform-specific tool configurations |
src/copilot-cli/*.agent.md (18 files) |
Copilot CLI agent definitions regenerated with name field and shell tools |
src/claude/*.md (18 files) |
Claude Code agent definitions regenerated with model specifications and Claude-specific tool references |
|
Caution Review failedThe pull request is closed. Note Other AI code review bot(s) detectedCodeRabbit has detected other AI code review bot(s) in this pull request and will avoid duplicating their findings in the review comments. This may lead to a less comprehensive review. 📝 WalkthroughWalkthroughAdds standardized "Activation Profile" sections (Keywords + Summon) across agent/profile markdowns, expands many agent descriptions and templates, introduces Serena initialization and several .serena memory/policy artifacts, and removes the Get-AutoGeneratedHeader function/export from the build module. No runtime logic besides the module export change. Changes
Estimated code review effort🎯 3 (Moderate) | ⏱️ ~25 minutes Points needing special attention:
Possibly related PRs
Pre-merge checks and finishing touches✅ Passed checks (3 passed)
📜 Recent review detailsConfiguration used: Repository YAML (base), Organization UI (inherited) Review profile: CHILL Plan: Pro 📒 Files selected for processing (85)
Comment |
There was a problem hiding this comment.
Actionable comments posted: 0
Caution
Some comments are outside the diff and can’t be posted inline due to platform limitations.
⚠️ Outside diff range comments (12)
src/vs-code-agents/memory.agent.md (1)
1-6: Add AUTO-GENERATED FILE header comment—file lacks required safety marker.File was regenerated via Generate-Agents.ps1 and content is consistent with template. However, both
src/vs-code-agents/memory.agent.mdandsrc/copilot-cli/memory.agent.mdare missing the AUTO-GENERATED FILE header comment required by coding guidelines.Add this header to both files:
- Format:
# AUTO-GENERATED FILE - DO NOT EDIT DIRECTLY- Include source template reference and generator script name
- Place before the YAML frontmatter
src/copilot-cli/planner.agent.md (1)
2-17: Add missing AUTO-GENERATED FILE header comment.File is missing the required
# AUTO-GENERATED FILE - DO NOT EDIT DIRECTLYheader comment before YAML frontmatter. Per coding guidelines, generated agent files insrc/copilot-cli/must include this header with source template and regeneration command reference to prevent accidental manual edits.src/copilot-cli/devops.agent.md (1)
2-17: Add AUTO-GENERATED FILE header to generated agent file.Generated Copilot CLI agent files must include a header comment at the top (before YAML frontmatter) indicating they are auto-generated and should not be edited directly. Missing from lines 1-2. Add:
<!-- AUTO-GENERATED FILE - DO NOT EDIT DIRECTLY Source: templates/agents/devops.shared.md Generator: build/Generate-Agents.ps1 -->src/vs-code-agents/task-generator.agent.md (1)
2-17: Add required AUTO-GENERATED FILE header to generated file.This file is generated from
templates/agents/task-generator.shared.mdand must include a header comment per coding guidelines. Add the following header comment at the top of the file before the YAML frontmatter:# AUTO-GENERATED FILE - DO NOT EDIT DIRECTLY # Source: templates/agents/task-generator.shared.md # Generator: build/Generate-Agents.ps1Generated agent files in
src/vs-code-agents/must never be edited manually. Regenerate by running the build script, which should automatically add this header.src/vs-code-agents/planner.agent.md (1)
2-17: VS Code agent has incorrect model field.The template (
templates/agents/planner.shared.md) does not specify a model, andsrc/copilot-cli/planner.agent.mdcorrectly has no model field. However,src/vs-code-agents/planner.agent.mdline 5 specifiesmodel: Claude Opus 4.5 (anthropic)whilesrc/claude/planner.mdline 4 usesmodel: sonnet. Remove the model field from the VS Code agent to match the template structure, or change it tomodel: sonnetto align with the Claude source.templates/agents/explainer.shared.md (1)
90-110: Replace absolute path examples with placeholder syntax in anti-pattern section.Lines 96, 99–100 contain literal absolute paths that violate coding guidelines. Use placeholder syntax instead:
[DRIVE]:\Users\[username]\repo\docs\guide.mdfor Windows/Users/[username]/repo/docs/guide.mdfor macOS/home/[username]/repo/docs/guide.mdfor LinuxThis preserves the anti-pattern examples while passing validation.
src/claude/explainer.md (1)
95-101: Replace absolute paths in anti-pattern examples with placeholder syntax.Lines 107-110 show forbidden absolute paths (Windows C:, macOS /Users/, Linux /home/) as examples of what not to do. Coding guidelines prohibit absolute paths in documentation without exception. Replace with placeholders to demonstrate the anti-pattern without violating the requirement.
Use syntax like
[WINDOWS_DRIVE]:\Users\[USERNAME]\repo\docs\guide.mdand/Users/[USERNAME]/repo/docs/guide.mdto preserve the teaching value while complying with the relative path requirement.src/vs-code-agents/explainer.agent.md (1)
94-101: Replace absolute paths in anti-pattern examples with placeholder syntax.The Path Normalization Protocol section shows forbidden absolute paths. Coding guidelines prohibit absolute paths in documentation—including in anti-pattern examples. Use placeholder syntax to demonstrate the anti-pattern without breaking the rule.
Fix the source template
templates/agents/explainer.shared.md, then regenerate all derived files. Replace paths with placeholders:
- Windows:
[WINDOWS_DRIVE]:\Users\[USERNAME]\repo\docs\guide.md- macOS:
/Users/[USERNAME]/repo/docs/guide.md- Linux:
/home/[USERNAME]/repo/docs/guide.mdsrc/claude/pr-comment-responder.md (1)
638-643: Missing formal Execution Mindset section. Other agent templates (analyst, critic, roadmap, etc.) include an explicit "Execution Mindset" section with Think/Act pattern. PR Comment Responder should include this for consistency and clarity of operational behavior.Consider adding:
## Execution Mindset **Think:** "Every comment deserves acknowledgment and systematic resolution" **Act:** Triage by actionability, track conversations, delegate appropriately **Verify:** No comment left unaddressed, no reviewer ignoredSuggested location: Between "Communication Guidelines" and "Bot-Specific Handling" sections.
src/copilot-cli/explainer.agent.md (1)
92-110: Pipeline failure: Path normalization violations in anti-pattern examples need resolution.Lines 96, 99, 100 show absolute paths marked as "FORBIDDEN" examples. Pipeline validation correctly flags these as violations, but they're presented as demonstrative anti-patterns. Two options: (1) convert anti-pattern paths to
[example-path]placeholders, or (2) move anti-patterns to a separate callout where linting is suppressed. Verify the approach with your CI configuration.templates/agents/pr-comment-responder.shared.md (1)
656-682: Duplicate Handoff sections must be removed.Lines 656–663 ("## Handoff") and lines 675–682 ("## Handoff Protocol") contain identical content. Per coding guidelines, keep only one Handoff Options section in agent templates.
🔎 Remove the duplicate section:
## Handoff This agent primarily delegates to **orchestrator**. Direct handoffs: | Target | When | Purpose | |--------|------|---------| | **orchestrator** | Each comment analysis | Full workflow determination | | **orchestrator** | Each implementation | Code changes | -## Handoff Protocol - -This agent primarily delegates to **orchestrator**. Direct handoffs: - -| Target | When | Purpose | -|--------|------|---------| -| **orchestrator** | Each comment analysis | Full workflow determination | -| **orchestrator** | Each implementation | Code changes |src/vs-code-agents/pr-comment-responder.agent.md (1)
656-682: Duplicate Handoff sections inherited from template.Lines 656–663 and 675–682 duplicate the Handoff content. This issue originates in the template (
templates/agents/pr-comment-responder.shared.md). Fix the template, then regenerate this file with.\build\Generate-Agents.ps1.🔎 Fix the template and regenerate:
Edit
templates/agents/pr-comment-responder.shared.mdto remove lines 675–682 (the duplicate "## Handoff Protocol" section), then run:.\build\Generate-Agents.ps1This will regenerate
src/vs-code-agents/pr-comment-responder.agent.mdwith the corrected content.
🧹 Nitpick comments (1)
src/claude/implementer.md (1)
17-17: Minor wording suggestion for Summon statement.Line 17 uses "hard to test" twice in close proximity: "If it's hard to test, flag it—that reveals deeper design problems." Consider rephrasing the second clause to strengthen specificity, e.g., "that signals deeper design flaws" or "that uncovers structural issues."
Add activation vocabulary and summoning prompts to all 18 agents across all platforms (Claude, VSCode, Copilot CLI) to improve model's ability to select the right specialist agent immediately. Changes per agent: - Updated description field with activation-focused language - Added Activation Profile section with: - Keywords: 20-word activation vocabulary cluster - Summon: 4-sentence character description for invoking specialist Files modified: - 18 Claude agents (src/claude/) - 18 VSCode agents (src/vs-code-agents/) - 18 Copilot CLI agents (src/copilot-cli/) - 18 shared templates (templates/agents/) Also removed dead code: - Removed Get-AutoGeneratedHeader function from Generate-Agents.Common.psm1 (unused function that could accidentally add unwanted headers to files) 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
96e36b6 to
00920b6
Compare
BREAKING: All Claude Code, VS Code, and Copilot sessions MUST initialize Serena before any other work. Changes: - CLAUDE.md: Added mandatory Serena init section at top - AGENTS.md: Added mandatory Serena init section at top - .github/copilot-instructions.md: Added mandatory Serena init section Also includes: - Retrospective analysis of session failures (.agents/retrospective/) - Skills and user preferences in Serena memory (.serena/memories/) Root cause: Not initializing Serena led to missing project context, repeated mistakes, and ignoring user preferences stored in memory. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Pull Request
Summary
Add activation vocabulary and summoning prompts to all 18 agents across all platforms to improve the model's ability to select the right specialist agent immediately.
Changes
descriptionfield with activation-focused language in YAML frontmatter## Activation Profilesection to each agent with:templates/agents/Generate-Agents.ps1Type of Change
Testing
Agent Review
Security Review
.agents/security/)Other Agent Reviews
Checklist
Related Issues
🤖 Generated with Claude Code
Note
Adds activation profiles across all agents/templates, mandates Serena MCP initialization in docs, removes the auto-generated header utility, and persists retrospective learnings to Serena memory.
descriptionand new## Activation Profile(Keywords + Summon) across all agent definitions insrc/claude/*,src/copilot-cli/*,src/vs-code-agents/*, andtemplates/agents/*.AGENTS.md,CLAUDE.md, and.github/copilot-instructions.md.Get-AutoGeneratedHeaderand its export frombuild/Generate-Agents.Common.psm1..agents/retrospective/..serena/memories/*(session failures, new skills, and user preference: no auto-generated headers).Written by Cursor Bugbot for commit 4ca9397. This will update automatically on new commits. Configure here.