Claude Agentic

Plugin claude-agentic in vanssata/ai-riskrail. Commands read from the repo's manifest — 1 month ago.

Skills (10)

• ai-audit

  Audit a repository against the twelve plays of the AI-native SDLC playbook and produce docs/ai-sdlc-adoption-plan.md — inventory, 0–3 gap scores with citations, three phased adoption phases respecting the dependency order, guardrail fixes and open questions. Read-only except for that one file. Use o

• ai-init

  Survey an existing production codebase and build the .ai/ agentic engineering infrastructure around it — knowledge base, policies, risk tiers, workflows, agent contracts — then write an initial assessment. Read-only with respect to application code. Use on "set up agentic infrastructure", "analyse t

• ai-status

  Show where the current agentic task stands — which .ai/ root governs this directory, id, stage, next action, risk tier, current step and its allowed files, test/e2e/review/security status, open risks, what the always-loaded instruction block costs — plus what each tier resolves to on this plan, the

• ai-task

  Run one task through the agentic pipeline with the gates its risk tier requires. In the default solo profile T0–T3 run in direct mode (name the files, edit, one verification run at the end, failures fixed as one batch, one BALANCED-tier review at T2; at T3 the plan is written in plan mode and approv

• project-init

  Scaffold a repo for Claude Code and/or Codex and the AI-Native SDLC flow — creates docs/sdlc/ (intent/specs/plans/adr with templates), the runtime directory (.claude/{settings.json,memory,plans,agents,skills} and/or .codex/{config.toml,memory,plans,agents,skills}), .gitignore entries and an instruct

• project-update

  Bring an existing project's claude-agentic files up to date with the installed plugin — .ai/ policies, workflows, agent contracts and templates, the managed block in CLAUDE.md, the docs/sdlc templates and .gitignore entries. Untouched files are replaced, edited files are three-way merged keeping eve

• sdlc-intent

  SDLC stage 1 — turn a rough idea into docs/sdlc/intent/<slug>.md (problem, outcome, affected users & systems, constraints, open questions) through a short one-question-at-a-time brainstorm. Use when starting a feature, "write an intent", "let's define what we want", before any spec or code.

• sdlc-plan

  SDLC stage 3 — from an intent + spec produce docs/sdlc/plans/<slug>.md, an ordered implementation plan (files that change, order of work, risks, proof/tests, rollback) that the main session then executes step by step. Use after /sdlc-spec, "plan the implementation", "make the plan from the spec".

• sdlc-spec

  SDLC stage 2 — read an intent file and produce docs/sdlc/specs/<slug>.md, a requirements and design spec that applies the project's CLAUDE.md conventions and available skills as organisational policy, and ends with explicitly flagged concerns (especially contradicting policies). Use after /sdlc-inte

• usage-report

  Report agent token usage and cost from local transcripts at zero model cost, for Claude Code and for Codex. Use when the user asks what a session/day/project cost, how many tokens were burned, which model spent the most, or how the two runtimes compare — instead of parsing transcripts with the model

Similar plugins