Major polish, library implementation, claude plugin implementation #6
Conversation
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
## Changes
- Streamline package structure (src/ with prompts/, tests/)
- Extract prompts to editable markdown files (src/prompts/*.md)
- Fix CI pipeline (proper package installation with uv)
- Remove incomplete terminal-bench
- Clean up obsolete documentation files
- Add build artifacts to .gitignore
## Package Structure
```
src/
__init__.py # Public API exports
api.py # Core execute() function
complexity.py # Complexity estimation
config.py # Configuration
hooks.py # SDK hooks for quality
prompts/
__init__.py # Prompt loader
executor.md # Executor agent prompt
reviewer.md # Reviewer agent prompt
planner.md # Planner agent prompt
complexity.md # Complexity estimation prompt
```
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
edgarpavlovsky
force-pushed
the
e/polish-library-plugin
branch
from
168f736 to
0e2d925
Compare
## Changes - Add execute->review loop for MODERATE mode (until >95% completion) - Add plan->execute->parallel_reviews loop for FULL mode (2/3 majority) - Remove SIMPLE mode (merged into SINGLE_TURN) - Extract data models to src/models.py - Add loop implementations in src/loops.py - Add prompt builder in src/prompts/builder.py - Update reviewer prompt with ISSUES format for structured feedback ## Architecture SINGLE_TURN: Direct Opus call (trivial tasks) MODERATE: while !complete: execute() -> review() FULL: plan() then while !complete: execute() -> 3_parallel_reviews() ## Key Features - Feedback from reviews flows to next execution iteration - Parallel reviews with asyncio.gather() in FULL mode - Majority (2/3) voting for FULL mode completion - Max 5 iterations to prevent infinite loops - Configurable via max_iterations parameter Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Update README with new execution modes and loop behavior - Update complexity prompt to explain what happens after each level - Remove references to deprecated SIMPLE mode - Add iterations field to ExecutionResult docs - Update project structure to include new files Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Change max_iterations default from 5 to None (infinite) - Add FIRETEAM_MAX_ITERATIONS env var for global config - Create Claude Code plugin command with config docs - Update loops.py to use itertools.count() for infinite iteration - Update README with new defaults and config options Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
edgarpavlovsky
commented
Jan 9, 2026
src/api.py
Outdated
| return ExecutionResult(success=False, mode=mode, error=f"Unknown mode: {mode}") | ||
|
|
||
|
|
||
| async def _single_turn( |
Member
Author
There was a problem hiding this comment.
let's move _single_turn() somewhere else - we can be more modular here
edgarpavlovsky
commented
Jan 9, 2026
src/complexity.py
Outdated
| COMPLEX = "complex" # Full Planner + Executor + triple Reviewer | ||
|
|
||
|
|
||
| async def estimate_complexity(goal: str, context: str = "") -> ComplexityLevel: |
Member
Author
There was a problem hiding this comment.
we should pass in the existing codebase here for better estimation
…lexity estimation 1. Move single_turn() from api.py to loops.py for better organization - All execution implementations now live in loops.py - api.py is now a thin dispatch layer 2. Add codebase exploration to complexity estimation - New project_dir parameter enables Claude to explore the codebase - Uses read-only tools (Glob, Grep, Read) in plan mode - More accurate complexity estimates for existing projects - Backward compatible - no project_dir = quick estimation Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.
This pull request makes significant updates to documentation, configuration, and CI/CD workflows for the Fireteam project. It removes older rule and environment files, introduces new plugin and agent documentation, updates the GitHub Actions workflow to streamline and clarify test execution, and removes legacy test suite summaries. The changes focus on modernizing project setup, improving clarity for contributors, and ensuring best practices for autonomous agent development and testing.
Documentation and Rules Modernization
.ai-rules.mdand.cursorrules, consolidating project setup guidance. [1] [2].claude-plugin/plugin.jsonand.claude-plugin/commands/fireteam.mdto document the Fireteam plugin, its configuration, and usage. [1] [2].claude/CLAUDE.mdoutlining agent principles for testing, quality gates, escalation, and code quality.CI/CD Workflow Improvements
.github/workflows/test.ymlto clarify test jobs, remove Node.js/Claude CLI setup, and unify dependency installation and test markers. Integration tests job removed; now only fast and integration tests run. [1] [2] [3] [4] [5]Environment and Test Suite Cleanup
.env.exampleand outdated test suite summary files (TEST_SUITE_SUMMARY.md,TESTING_COMPLETE.md) to reduce confusion and duplication. [1] [2] [3]TODO.mdto encourage future improvements to be tracked elsewhere.