Generate evidence-based implementation plans for multi-file Claude Code changes.
npx skills add https://github.com/anilcancakir/claude-code --skill ac-plan-anilcancakir
Please help me install this Agent Skill. Skill: /ac:plan Source: https://github.com/anilcancakir/claude-code/tree/main/plugins/ac/skills/plan Command: npx skills add https://github.com/anilcancakir/claude-code --skill ac-plan-anilcancakir
Producing a high-quality implementation plan across multiple files and decisions—grounded in codebase evidence—without rushing uncertainties or writing an unexecutable spec.
Ask the planner by giving a topic like: /ac:plan add an endpoint to create and list tasks with validation.
High-intent search queries and answers about installing and using this skill.
An implementation plan for multi-file changes is generated by surveying the repository, running parallel research subagents, and conducting an interactive decision-tree interview. This produces tier-assigned planning artifacts grounded in actual codebase evidence.
Evidence-based planning for cross-module refactoring grounds your strategy in actual codebase behavior by reading referenced code directly and auditing for reuse, quality, and efficiency. It ensures load-bearing decisions are validated through an interactive interview process.
Yes, you can chain planning into code execution automatically by enabling an auto mode flag. After the tier-assigned plan is delivered, the workflow chains into execution while gating only workflow recovery gates automatically.
To write a spec from a free-form topic for non-trivial feature work, you initiate an interactive interview that walks you through every load-bearing decision with a recommended-first branching style. The system then audits for reuse and efficiency to produce a structured plan.
Yes, interactive spec writing works with existing YAML task specifications by accepting them as an alternative input source alongside free-form topics. The planner reads the YAML specification to drive the interview and generate tier-assigned plan artifacts.
You should avoid automated tiered planning for feature development when your changes are trivial or confined to a single file without cross-module dependencies. The interactive interview and parallel research subagents are designed specifically for non-trivial feature work and complex refactors.
iurykriegerCentralize vault write operations with phase-based workflows and bidirectional linking.
gaiastudio-aiIdentify performance bottlenecks in changed files during story reviews.
supertyrelleCompare two technical approaches and produce a structured tradeoff analysis with a recommendation.
stark-ai-deInterview users to capture requirements and draft Codex-executable implementation specs.
