seal-harness
Official@seal-harness
Seal Harness provides a quality-gated development orchestration system with adversarial review gates, PR lifecycle management, and Haskell expertise.
Agent Skills by seal-harness
Showing 18 vetted skills indexed across 1 GitHub repositories.
plan-review-gate
Validates implementation plans using three parallel adversarial reviewers before user presentation.
create-issue
Generates structured GitHub issues with TDD plans, acceptance criteria, and agent instructions.
handoff
Writes a self-contained markdown handoff document so a fresh agent can resume work with full context.
handling-pr-comments
Automates triage, response, and resolution of GitHub pull request review comments.
pr-shepherd
Monitors GitHub pull requests through CI failures and review comments until merge.
setup
Detects project stack and configures Seal Harness instruction files, coverage thresholds, and command shims.
todo-md-maintenance
Adds, updates, and reads items in a project TODO.md file.
start
Orchestrates multi-agent workflows to autonomously process GitHub Issues from research through merged PR.
status
Generates a diagnostic report of Seal Harness installation, project setup, and configuration issues.
haskell-reviewer
Reviews Haskell code for correctness, idiomatic style, performance pitfalls, and best practices.
external-tools
Delegate implementation and review tasks to external AI CLI tools with cross-model adversarial review.
start-task
Assesses task complexity and routes work through simple or complex development workflows.
haskell-coder
Guides Haskell development covering type-driven design, Cabal builds, testing, and performance optimization.
brainstorming-extension
Enforces a five-agent design review gate after brainstorming before implementation planning begins.
design-review-gate
Runs a five-agent parallel review gate on design documents before implementation begins.
self-reflect
Extract reusable learnings from PR reviews and session patterns into a knowledge base.
orchestrated-execution
Executes work units through a 4-phase loop of implement, validate, adversarial review, and commit.
visual-review
Capture screenshots of web pages and UI with Playwright for visual inspection.
Frequently Asked Questions About seal-harness
FAQPage SchemaWhat tasks can I accomplish with Seal Harness skills?▼
You can run adversarial plan and design review gates, create GitHub issues with TDD plans, shepherd PRs through CI failures and review threads to merge, maintain TODO.md files, write and review Haskell code, capture Playwright screenshots for visual review, and generate session handoff documents.
Who is Seal Harness designed for?▼
Seal Harness targets developers running agent-assisted, tracked development work — particularly Haskell engineers and teams that want enforced quality gates, structured PR lifecycle management, and cross-model review via external Codex or Gemini delegation.
How do I install and start using Seal Harness?▼
Run the setup skill, which interactively detects your project, configures Seal Harness, and writes project-local files. Then invoke the start skill when beginning tracked work, and use the status skill anytime to diagnose installation state and project setup issues.
What are the prerequisites and dependencies for Seal Harness?▼
Skills assume a GitHub-based project with a docs/ directory for knowledge priming. Visual review requires Playwright for screenshots, external delegation requires Codex or Gemini CLIs, and Haskell skills expect GHC, Cabal, and optionally Nix build environments.
How do the review gates in Seal Harness work?▼
The plan-review-gate spawns three independent parallel reviewers after any plan is drafted, all of which must pass. The design-review-gate spawns PM, Architect, Designer, Security, and CTO reviewers after brainstorming, iterating until every role approves the design document.