Works with the AI Agents you already use
Claude Code
Cursor
Codex CLI
Muse
Dots+22 moreClaude CodeCursorCodex CLIMuseDots
OpenClaw
Manus
Grok
Grok Bot+18 agentsAgent Factory
A multi-agent software engineering suite that automates the full lifecycle from product shaping to PR publication.
$5
Agent Factory
Example session with this skill installed
Run the do-work lifecycle for the feature request in docs/feature-x.md. Map the codebase, design the solution, and prepare the construction plan.
- Read your context and instructions
- Compiled the agent factory
I have initiated the Agent Factory do-work lifecycle for the feature described in docs/feature-x.md. The repository has been mapped, and a technical blueprint has been generated.
Technical Blueprint: Feature-X Implementation
| Artifact ID | Baseline SHA | Change Surface | Verification Method |
|---|---|---|---|
| BP-2024-001 | 8f3d2a1 | src/auth/logic.ts | Unit tests + Integration |
| BP-2024-001 | 8f3d2a1 | src/api/routes.ts | Runtime API verification |
| BP-2024-001 | 8f3d2a1 | schema/user.sql | Migration dry-run |
Architecture Mapping
- Identity Provider: Integrates with existing
AuthServiceinpackages/core. - Data Flow:
POST /v1/feature-x->ValidationPipe->FeatureController. - Security Boundary: Requires
scope:writeon the JWT payload.
Acceptance Criteria Verification
- AC1: User can toggle Feature-X via the dashboard. (Status: Planned in
ui-components) - AC2: API returns 403 if user is not in the 'beta' group. (Status: Planned in
auth-middleware) - AC3: All changes are logged to the audit trail. (Status: Planned in
persistence-layer)
Next steps
- Review the generated Technical Blueprint for architectural alignment.
- Approve the plan to transition into the
construct-workphase. - Verify the baseline SHA matches your local main branch.
Connects securely to your tools. The creator never sees your data.
About this skill
Agent Factory is a comprehensive suite of 15 specialized skills designed to automate the entire software development lifecycle, from product discovery to CI monitoring. It solves the problem of agentic "drifting" where a single prompt tries to handle planning, coding, and testing simultaneously, leading to regressions and scope creep. By enforcing strict role boundaries, the Factory ensures that the code writer never reviews their own work and the tester never edits production code.
The package includes high-fidelity orchestrators like do-work and shape-work that manage complex lifecycles while delegating specific tasks to bounded specialists. These specialists handle codebase mapping, technical blueprinting, security auditing, and runtime QA verification. Each skill operates within a delegated Git worktree, ensuring all actions are commit-addressed and traceable.
What it does
- Autonomous Orchestration manages the full lifecycle from ticket intake to PR publication and CI monitoring.
- Product Shaping transforms vague ideas into evidence-backed, dependency-ordered work items.
- Independent Verification performs read-only security reviews, quality audits, and runtime QA without editing source.
- Technical Blueprinting creates versioned implementation plans before a single line of production code is written.
- Repository Mapping builds evidence-led maps of call trees, schemas, and integration boundaries.
How it works
- Initialize a project by providing a Git repository and a work item like a PRD or Jira ticket.
- Shape the work using specialized discovery paths like Lean experiments or Dual-Track discovery.
- Plan the implementation by generating a versioned technical blueprint mapped to acceptance criteria.
- Construct the solution in a dedicated worktree followed by independent security and quality gates.
Frameworks & tools
This skill set is built for Git-based workflows, specifically optimized for GitHub and GitLab environments. It integrates with standard development tools like rg, diff, and Mermaid for visualization.
Why this beats prompting it yourself
Generic prompts often mix concerns, causing agents to skip tests or ignore security risks to make code "just work." Agent Factory enforces a multi-agent "separation of powers" where the writer, reviewer, and tester are distinct roles with immutable boundaries.
Use cases
- Automating the conversion of PRDs into sliced, testable GitHub issues.
- Conducting rigorous, read-only security audits on existing pull requests.
- Implementing complex features across shared abstractions with automated regression testing.
- Monitoring CI pipelines and summarizing review feedback for human developers.
Known limitations
Requires a functional Git environment and a defined worktree structure. It does not support direct deployment to production or merging PRs without human intervention.
What's included (15 skills)
author-tests
Create or update focused automated tests and fixtures for an immutable construction checkpoint, run configured verification commands, and report production defects without editing production code. Use in parallel with the security review after construction.
challenge-product
Independently challenge a product framing, PRD, or opportunity by testing assumptions, alternatives, risks, metrics, and experiment logic. Use only when delegated by the shape-work parent as a read-only disconfirming review.
construct-work
Implement an explicitly approved software plan or a bounded remediation request as the sole production-code writer. Use after do-work delegates construction with fixed scope, acceptance criteria, baseline revision, and checkpoint requirements.
design-solution
Turn repository evidence and approved decisions into a versioned technical blueprint without writing production code.
do-work
Orchestrate a software work item from a ticket, PRD, specification, URL, or free-form request through planning, construction, independent verification, pull or merge request publication, CI monitoring, a visual show-me explanation, and human handoff. Use when the user explicitly asks to run the complete Agent Factory lifecycle.
map-codebase
Build a read-only, evidence-led map of the repository before solution design or construction.
research-product
Gather bounded, source-backed user, product, market, domain, repository, and feasibility evidence for a product-shaping question. Use only when delegated by the shape-work parent; never make product decisions or contact research participants.
review-code-quality
Run an independent, exceptionally strict read-only review of a tested change for specification fidelity, correctness, structural simplicity, maintainability, abstraction quality, and codebase health. Use as the final local quality gate before publication.
review-security
Perform an independent read-only security review of the changed production attack surface at a pinned Git revision. Use after construction to find and evidence every validated security risk without modifying code.
review-technical-plan
Independently review a technical blueprint for completeness, risk, traceability, and unresolved material decisions before construction.
review-work-items
Independently review proposed product work items for evidence traceability, outcome orientation, vertical slicing, dependency integrity, testability, risk coverage, and readiness for do-work. Use only when delegated by the shape-work parent.
shape-work
Turn a rough product idea, customer problem, research packet, opportunity, or complete PRD into one or more evidence-labelled, dependency-ordered software work items ready for the Agent Factory do-work lifecycle. Use only when the user explicitly asks to discover, shape, validate, de-risk, slice, or prepare product work.
show-me
Help the user understand the current topic visually with concise diagrams, code-shape sketches, and focused HTML artifacts.
verify-qa
Independently verify approved acceptance criteria against the tested Git revision using runtime evidence for observable application, API, CLI, or library behavior. Use after security and automated testing complete.
watch-change
Monitor an existing GitHub pull request or GitLab merge request for CI completion and review feedback, returning structured observations to the do-work parent without changing code, merging, or deploying. Use only after a locally verified change is published.
How to use
Skills on Agensi run inside your AI agent. Copy the prompt, paste it into your agent, and it connects to Agensi and installs this skill for you. If a payment is needed, your agent shows it to you first.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill passes our trust checks below.
Security scanned
Passed our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scanned
Verified clean 1 day ago
- Passed all security checks, Safe to install
Requires
Python
Git