jaydubya818/software-factory-skills
Cross-harness delivery, review, Agent-Native, visual, and software-factory skills for production work.
Record approved decisions as concise architecture decision records.
Audit, monitor, compare, or repair another agent's session or PR when asked.
Operate granted Agent-Native apps through Dispatch MCP; use for /an.
Design API contracts, compatibility, and duplicate-safe mutations.
Use when a non-trivial change needs types and module boundaries designed before code.
Use when an important artifact needs independent solutions compared and synthesized.
Use when the user wants working preferences captured in a reusable mode skill.
Use when an authorized pull request needs ongoing CI and review follow-through.
Use when attaching screenshots or recordings to a GitHub PR.
Use when checking measured benchmark claims.
Use when a change needs downstream risks traced and its key safety assumption proved.
Use when the previous technical answer needs a direct, jargon-free restatement.
Create or change CI/CD pipelines, quality gates, and delivery controls.
Separate orchestration from shared mechanics while preserving transactions.
Use for /correct or requested fixes to repeated agent mistakes.
Use when creating a project-local formal verifier.
Prepare least-privilege PostgreSQL agent access when requested.
Use when a repository needs a repeatable driver that proves real UI, CLI, or service behavior.
Retire APIs or migrate consumers and persistent data across releases.
Use when code needs generated clutter removed and structure simplified without behavior changes.
Delegate bounded independent work while Claude Fable retains judgment.
Delegate costly independent work while a frontier model retains judgment.
Capture runtime evidence with video, screenshots, or measured output.
Configure governed delivery sources, schedules, isolation, and action gates.
Monitor one authorized PR with separate fix, publish, reply, and merge gates.
Triage configured product signals under separate fix, reply, and close policies.
Read-only summary of configured work needing human decisions.
Audit signal history for recurring problems and systemic fixes.
Resume eligible interrupted runs without repeating completed actions.
Triage PR queues under separate reply, approval, and merge gates.
Ship authorized work under separate publish, merge, deploy, and notification gates.
Monitor authorized delivery tasks for verified stalls.
Use when novel or multi-phase work needs an auditable workflow and no narrower playbook fits.
Use when failing pull-request checks need a root-cause fix and a verified rerun.
Use when a branch needs conflicts resolved and the merged result verified.
Build complete, accessible UI flows with explicit states and browser checks.
Use when pull-request feedback needs to be fetched and summarized for action.
Run bounded Greptile review-and-fix cycles on a PR, MR, or shelved change. Use when Greptile is an explicit delivery gate.
Run greploop through @greptile-apps when the normal trigger rejects a large PR. Use only with the companion greploop skill.
Write verified session context when asked to hand work to another agent.
Use for runtime, ownership, layering, code-walkthrough, and placement questions.
Use when asked to restate the user's goal.
Use when a change's assumptions, evidence, or design need adversarial review.
Use for /ketchup or a chat progress catch-up.
Use when refreshing a project's formal verification.
Use when a project's verification driver or feature map may be stale.
Use when a correct pull request needs clearer history, description, or reviewer guidance.
Preserve evidence and authority when Mission Control governs delivery.
Isolate repository edits in a task-owned Git worktree. Skip read-only work.
Use when code comments need structural review, root-cause fixes, and constraint encoding.
Add or review logs, metrics, traces, and actionable alerts.
Investigate slow paths and optimize measured bottlenecks.
Compare, judge, or combine competing plans into one executable path.
Proceed through routine ambiguity when autonomous progress is requested.
Use for help choosing or using Jstack workflows.
Use when non-trivial work should follow the smallest rigorous Jstack workflow.
Apply when repeated fixes fail and a shared assumption may be wrong.
Apply when validation or adapters leak from external boundaries into core logic.
Apply when a rerunnable script, codemod, generator, or check makes work auditable.
Apply when a repeated instruction should become enforceable in tests, lint, metadata, or tools.
Apply when a novel design needs concrete candidates compared before selection.
Apply when scope threatens a polished, clear, and trustworthy user experience.
Apply when a measured result needs its limits explained.
Apply when a workaround would hide a defect that can be reproduced and traced.
Apply when core types, data structures, or shared-state boundaries will shape the system.
Apply when large files, logs, or broad exploration threaten reasoning focus.
Apply when a refactor is adding wrappers or abstractions instead of simplifying.
Apply when an operation must tolerate retries after partial completion.
Apply when replacing an internal API without leaving ambiguous dual paths.
Apply when layers, wrappers, or hidden state make simple behavior hard to follow.
Apply when repeated state conditionals reveal a missing domain model.
Apply when safe reversible implementation details can proceed inside scope.
Apply when a bounded rewrite should converge without throwaway compatibility layers.
Apply when the actual candidate and observable result still need verification.
Apply when a requirement breaks a foundational assumption and needs coherent redesign.
Apply when concurrent actors share mutable files, branches, keys, or objects.
Apply when multi-step work must end each ordered unit in a verified state.
Apply when new work would otherwise build on dead or redundant structure.
Apply when tests mirror internals instead of asserting observable outcomes.
Apply when typed APIs or schemas should make invalid states unrepresentable.
Write a verified setup or fix prompt for a teammate's agent on request.
Use a red, yellow, or green status line for compact agent completion reports.
Read primary docs for external APIs, packages, or high-stakes behavior.
Use when resuming work or reconstructing project context from workspace-scoped history.
Use when approved session lessons should become durable workflow improvements.
Retrieve local Clips Rewind context for something the user recently saw or said.
Validate assumptions behind changes to billing, API data, or defaults.
Review security or harden sensitive trust boundaries with abuse-case tests.
Use when the user wants to configure Jstack's per-role Claude models.
Use for /sharpen to clarify a task before work.
Use when long-running work needs an auditable decision and evidence log.
Check host usage limits between bounded waves of long-running work.
Use when authorized independent workers can run without shared mutable state.
Use when TDD is requested or a bug has a cheap, direct regression target.
Use when runtime behavior and design rationale should become one clear mental model.
Use when docs, RFCs, READMEs, PRs, or commits need structured and direct prose.
Use when the user requests an intentionally strict maintainability review.
Use for a deep security and correctness diff review.
Use for combined correctness and code-quality reviews.
Build a verified Agent-Native app from a workflow, skill, or visible artifact.
Use when TypeScript needs focused type-safety, API, or error-handling guidance.
Remove filler and generic AI phrasing from human-facing text. Use before sending or committing prose you edited.
Edit or review a running local app through Agent-Native Design.
Create an interactive Agent-Native plan for visual review.
Create an Agent-Native visual recap from a substantial code change.
Operate a web app through its page MCP tools; use for /webmcp.
Use when authored work needs a concise summary for a concrete time window.
Use when design rationale must be recovered from source history and available records.