Loki Mode
Deploy products from specs autonomously with multi-agent coordination
What You Can Do
Activate Loki Mode to orchestrate a multi-agent team that autonomously takes any product specification—whether a Markdown PRD, GitHub issue, OpenAPI doc, or Jira ticket—and drives it through to deployment. The system makes decisions independently, coordinates across multiple AI providers (Claude, Codex, Gemini, Cline, Aider), and tracks progress through phases while maintaining state and handling failures gracefully.
Features
accepts PRDs, GitHub issues, OpenAPI docs, Jira tickets, or any structured specification format
routes tasks across Claude, Codex, Gemini, Cline, and Aider with abstract model tiers and degraded mode support
operates without stopping to ask questions; makes calls independently within defined phases
maintains session context, task queues, orchestrator state, and memory indexes across execution
structures work into RARV-C phases (Requirement, Architecture, Review, Validation, Closure) with task generation and phase progression
supports both Bash and Bun runtimes with seamless fallback; migration in progress for performance
provides JSON-formatted status reports, stats, and doctor diagnostics for monitoring execution
Example Output
Example 1: GitHub Issue → Deployed Feature
Input: GitHub issue #342 requesting user authentication with OAuth2
Loki processes: Extracts requirements → Generates architecture → Creates implementation tasks → Writes code → Runs tests → Deploys to staging → Validates integration
Output: OAuth2 authentication feature live on staging, session log with phase completion timestamps
Example 2: OpenAPI Spec → Backend API
Input: OpenAPI 3.0 spec for payment processing REST API
Loki processes: Parses endpoints → Generates data models → Creates route handlers → Implements validation → Adds error handling → Deploys to production
Output: Fully functional payment API with Swagger docs and automated test coverage
Example 3: PRD → Full Product
Input: 5-page product requirements document for a real-time collaboration tool
Loki processes: Extracts user stories → Plans architecture → Coordinates frontend/backend teams → Manages dependencies → Validates requirements → Ships MVP
Output: Deployed product with documented decisions and phase completion report
What's Included
- `SKILL.md`: Complete Loki Mode instruction file with phase logic, decision trees, and agent behaviors
- `skills/00-index.md`: Module registry and phase-specific skill mappings
- `.loki/state/` templates: Orchestrator state schema, session tracking, and task queue formats
- `skills/providers.md`: Multi-provider routing configuration and degraded mode fallbacks
- `docs/architecture/ADR-001-runtime-migration.md`: Runtime migration guide (Bash → Bun)
Who It's For
- Engineering managers — delegate product spec-to-launch execution to autonomous agents
- Startup founders — rapidly prototype and deploy MVPs from PRDs without manual coordination
- DevOps engineers — automate infrastructure-as-spec workflows triggered by Jira/GitHub
- Platform teams — orchestrate cross-team delivery pipelines from requirements to production
- Solo developers — scale output by running parallel autonomous sprints from issue backlogs
Best For
- Rapid MVP deployment from product requirements or issue descriptions
- Multi-phase product development coordinated across architecture, implementation, testing, and deployment
- API-first development triggered by OpenAPI specs or async specifications
- Autonomous sprint execution driven by GitHub issues or Jira tickets
- Cross-provider orchestration when you need fallback AI models or heterogeneous agent teams
You might also like
You can conduct thorough, defensible skin assessments that integrate visual analysis, tactile evaluation, client history, and contraindication screening into a documented consultation protocol. This skill guides you through systematic evaluation of skin conditions, helps identify when physician referral is necessary, and supports informed treatment planning that protects both you and your clients while staying within your scope of practice.
by asklokesh
You can submit your cosmetics formulation concept and receive comprehensive analysis across three critical dimensions: ingredient compatibility (interactions, solubility, pH stability), regulatory compliance (EU, US, China restrictions and claim substantiation), and market positioning. The skill processes ingredient concentrations, functional claims, and target markets simultaneously to flag compatibility issues, regulatory gaps, and stability risks before costly development work begins.
by asklokesh
You can evaluate new retail opportunities with segment-specific strategies, prepare for retailer negotiations with buyer teams armed with realistic velocity assumptions, and forecast sales impact before committing inventory. The skill helps you analyze underperforming doors, compare ROI of shelf space investment across channels, and build annual retail strategy with clear door mix targets and success metrics.
by asklokesh
You'll transform raw QC analytical data into structured batch release decisions that satisfy regulatory requirements. Claude guides you through specification comparison, statistical deviation analysis, root cause tracing, and corrective action planning—turning instrumental results (HPLC, viscosity, pH, microbiology) into clear hold/release recommendations with documented justification for audits and inspections.
by asklokesh
Rapidly develop and balance cosmetic formulations by analyzing how actives, solvents, emulsifiers, and preservatives interact within your system. You can evaluate ingredient compatibility before ordering materials, diagnose why batches fail (separation, crystallization, potency loss), and optimize texture and performance while maintaining stability and regulatory compliance. This skill reduces iteration cycles by catching formulation conflicts early and providing data-driven adjustment recommendations.
by asklokesh
Claude analyzes your code diffs and generates standardized commit messages that follow the Conventional Commits specification. The skill automatically determines the correct commit type, scope, and description based on the changes you've made, ensuring your messages are parseable by automation tools while remaining human-readable for code reviewers.
You get a rapid-response diagnostic tool that analyzes build logs, detects root causes of failures, and identifies performance bottlenecks in your CI/CD pipelines. Claude identifies breaking changes, flaky tests, configuration issues, and resource constraints—then proposes specific optimizations to reduce build times and improve deployment reliability.
You can systematically evaluate multi-channel DTC cosmetics campaigns to identify optimization opportunities, calculate true customer acquisition costs (CAC) vs. customer lifetime value (CLV) ratios, and pinpoint underperforming assets. This skill helps you move beyond vanity metrics to diagnose root causes of underperformance, segment performance by customer cohort, and recommend specific optimizations that increase ROAS and profitability across paid social, email, search, and influencer channels.
by asklokesh