Agentic Software Engineering

Plugin agentic-software-engineering in MarkCodering/DevKits. Commands read from the repo's manifest — 1 month ago.

Skills (11)

• adopt-agentic-engineering

  Assess and roll out agentic software engineering at team or organization level using positive-yield metrics, risk-adjusted autonomy, role and governance design, a five-level maturity model, and staged 30-day or 90-day adoption programs. Use for operating-model design, pilot planning, agent platform

• apply-delivery-playbook

  Select and apply a TRACE delivery playbook for a feature, bug fix, refactor, database migration, API integration, frontend/UX change, infrastructure/DevOps change, or AI/model-backed feature. Use when task type changes the required predicates, sequencing, evidence, rollout, or recovery controls.

• design-agent-route

  Inspect a real repository and convert an accepted Target Contract into a reviewable Design Scaffold, dependency DAG, bounded write surfaces, proof plan, risk controls, and human decisions. Use before non-trivial implementation, multi-agent execution, cross-component changes, migrations, or work wher

• distill-context-capsule

  Compress ongoing engineering work into a source-linked Context Capsule for a fresh agent or human, preserving authoritative Target/Route versions, task state, decisions, evidence pointers, negative knowledge, risks, invalidation rules, and next action. Use at handoffs, context limits, stage boundari

• govern-agentic-release

  Build and evaluate an evidence-first merge and release decision with risk-tiered authority, supply-chain checks, canary or flag strategy, service predicates, observation window, rollback or forward-repair triggers, tested recovery, and evidence lineage. Use before merging or releasing agent-produced

• orchestrate-parallel-agents

  Design and run safe parallel coding-agent work as an isolated dependency DAG with readiness checks, worktree or environment leases, bounded child authority, versioned artifacts, explicit integration ownership, fan-in verification, and review backpressure. Use when two or more independent tasks could

• ready-repository-for-agents

  Audit and improve a codebase as an agent engineering harness by making important truth executable, instructions local, feedback fast, errors informative, tools narrow, context retrievable, state external, and evaluation representative. Use when coding agents waste searches, miss commands or boundari

• run-bounded-implementation

  Execute one agent coding issue under an explicit objective, write boundary, budget, pass predicate, authority model, and typed stop conditions. Use when implementing a scoped DAG node, correcting a failed predicate, or preventing autonomous coding from drifting into retries, cleanup, new dependencie

• trace-agentic-delivery

  Coordinate end-to-end software work through the TRACE control loop: Target, Route, Assemble, Compress, and Evaluate. Use for substantial features, bug fixes, migrations, refactors, incidents, or multi-stage coding-agent work that needs explicit artifacts, gates, failure routing, and evidence-backed

• verify-agentic-delivery

  Independently verify an agent-produced candidate against its Target, accepted Route, exact diff, and risk tier using scope, static, behavioral, scenario, adversarial, test-quality, and operational evidence. Use for code review, pre-merge validation, failed checks, release decisions, or any correctne

• write-target-contract

  Convert a software request into a human-owned Target Contract with actionable actors and boundaries, concrete deliverables, layered acceptance predicates, non-goals, constraints, risk, authority, and recovery. Use before coding when requirements are vague, outcome correctness matters, or an agent ne

Similar plugins