feat Meta Agent and more#15
Merged
Merged
Conversation
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Introduce Meta-Agent, an agent architect that designs custom specialized agents on-the-fly using embedded prompt engineering best practices from docs/Prompt_Best_Guides.md. The Meta-Agent crafts tailored system prompts, selects minimal tool sets, executes tasks, and returns structured JSON results. - Add MetaAgent implementation with full tool access (meta.go) - Add MetaAgent system prompt embedding prompt engineering best practices - Integrate delegate_meta tool into ConductorAgent for task delegation - Update conductor prompt with Meta-Agent capability documentation - Add meta_max_steps configuration support Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
After Conductor delegates to Meta-Agent, the designed agent is now automatically parsed, registered as a permanent delegate tool, and described in the system prompt for future reuse. Also added the `finish` tool to tools.json and updated the conductor prompt to accurately describe the Meta-Agent workflow. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
When Meta-Agent LLM output doesn't follow the strict <agent_design>/ <execution_result> JSON format, the system now attempts heuristic extraction of agent name and tools via regex patterns, enabling custom agent registration even with non-compliant LLM responses. Falls back gracefully to raw output when extraction fails. Also added functional testing methodology documentation to CLAUDE.md. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Simplify the delegate_meta error path: when Meta-Agent output doesn't follow the strict <agent_design>/<execution_result> format, return the raw output directly instead of attempting heuristic extraction. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Add --disable-agents CLI flag to conditionally exclude sub-agent delegate tools from the Conductor, simplifying isolation and debugging of individual agents. Switch TUI input from single-line textinput to multi-line textarea. Also simplify CLAUDE.md and extract detailed testing methodology into docs/TESTING.md. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
…tion Replace the XML-based <agent_design>/<execution_result> blocks with a single JSON object containing thinking, agent_name, agent_design, tools_used, and result fields. Add extractJSONObject to handle markdown fences and surrounding text. On parse failure, retry up to 3 times with format correction feedback to the Meta-Agent. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Add meta_retry_count to [agent] config section (defaults to 5 when unconfigured). Replaces the hardcoded const maxRetries=3 with a ConductorAgent.metaRetryCount field set from config. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Add MANDATORY Phase 0: Task Classification & Agent Selection at the start of the workflow. The Conductor now first decides the execution strategy: chat → Chat-Agent, coding → Repo+Coding, specialized → design via Meta-Agent FIRST, existing custom agent → reuse directly. Clarify that working agents (coding, chat, custom) are peers, while repo and meta are support agents. Relax mandatory repo analysis for custom-agent tasks. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
…TTP section Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
![sourcery-ai[bot]](https://avatars.githubusercontent.com/in/48477?s=60&v=4){kind=small}
There was a problem hiding this comment.
Sorry @iohub, your pull request is larger than the review limit of 150000 diff characters
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
No description provided.