analyze-with-file_skill

Overview

This skill provides an interactive, document-driven analysis workflow that records the full evolution of understanding while exploring a codebase or topic inline. It enforces immediate decision and finding recording, supports multi-round Q&A, and can convert conclusions into executable tasks. Designed for serial, non-delegated analysis, it preserves traceability and reproducibility of analysis sessions.

How this skill works

It initializes a dated session folder and a discussion.md that captures context, dimensions, and an evolving discussion timeline. The skill performs inline exploration of the repository (grep/glob/read) and builds per-perspective findings (single or up to four serial perspectives), updating discussion.md after each round. Users iterate with multi-round interactive prompts; critical decisions, key findings, assumption changes, and scope adjustments are recorded immediately. Optionally, conclusions can be turned into task files and executed after pre-execution checks.

When to use it

• When you need a reproducible, auditable analysis of architecture, performance, security, or implementation concerns
• When multiple perspectives should be explored serially and synthesized (technical, architectural, security, performance)
• When you want a structured, multi-round interactive review with mandatory decision recording
• When converting analysis outcomes directly into executable tasks or issue-ready artifacts
• When working in a team that requires traceability of why and how analysis choices were made

Best practices

• Start each session with clear focus areas and limit perspectives to four or fewer to keep rounds manageable
• Use --depth=deep for comprehensive codebase scans and --yes for automated flows when scope is settled
• Record decisions immediately using the provided Decision Record Format to preserve rationale and traceability
• Include project metadata files (project-tech.json, project-guidelines.json) in the repo so exploration surfaces richer context
• Use Quick Execute only after resolving dependencies and reviewing pre-execution analysis to avoid unintended edits

Example use cases

• Investigate authentication architecture and capture trade-offs and mitigations with an audit trail
• Perform a performance bottleneck analysis across code and config, iterating until root causes are clear
• Run a security-focused, multi-perspective review that records findings and recommended fixes as tasks
• Synthesize multi-perspective findings into a final conclusions.json and generate TASK-*.json for agile tracking
• Resume a prior session to continue investigation without losing prior decision context