pwrl

Extract, classify, deduplicate, structure, and save learnings from code, commits, tasks, and documentation

Create structured implementation plans with three tiers (Fast/Standard/Deep). Pure skill pipeline orchestrator—no agent routing.

Review code changes through 4-phase micro-skill pipeline (scope, prepare, analyze, report)

Execute implementation work efficiently through 4-phase micro-skill pipeline

Verify repository state and confirm session completion before committing.

Create a clear session commit with state and next steps. Orchestrates checkpoint and commit micro-skills, optionally chains to pwrl-learnings.

Extract actionable learnings from code, commits, tasks, documentation, errors, and reviews.

Gather context, check learnings, validate domain, and bootstrap requirements for planning workflow.

Validate that reviewed code matches requirements and detects scope creep.

Slice implementation plans into granular, executable task files. Use when: breaking down plans from pwrl-plan; creating actionable tasks for pwrl-work; need detailed task files in docs/tasks/to-do.

Classify work input and extract foundational context for execution

Draft commit message, handle version bumps, and create the session commit.

Classify and prioritize learnings by type, domain, severity, and applicability.

Identify and merge duplicate or redundant learnings while preserving historical lineage.

Persist deduplicated learnings to permanent storage with backups and version control.

Structure classified learnings for persistent storage with metadata, indexing, and navigation.

Analyze code for quality, security, tests, documentation, and integration issues.

Prepare code review environment, gather artifacts, and configure analysis tools.

Generate comprehensive review report and determine final approval status.

Post review findings back to GitHub PR with comments, formal reviews, and labels. Final phase of pwrl-review orchestrator.

Comprehensive testing framework for micro-skill pipeline architecture validation, including unit, integration, compatibility tests, and consolidation audit.

Set up execution environment, create task lists, and select execution mode

Choose tier, render plan from templates, embed learnings, and save to docs/plans/.

Create technical design and implementation unit decomposition for planning workflow.

Perform local and external research to inform technical design decisions during planning.

Review and update existing learnings to keep knowledge current

Update docs/learnings/INDEX.md from current learning documents, typically after pwrl-end-session.

Execute tasks with quality gates (inline, serial, or parallel modes)

Review, simplify, and consolidate code after execution

Synchronize task status with GitHub Issues when integration is enabled

Enforce deterministic phase completion in PWRL workflows (review, work, plan, etc.). Use to: validate artifact generation, verify quality gates, block progression without completion. Works across all PWRL pipeline skills. Agent-agnostic, checkpoint-driven validation.