Agent-Skills.md

Agent Skills: Code Flow Analysis

>-

UncategorizedID: jagreehal/jagreehal-claude-skills/code-flow-analysis

Repository

jagreehal/jagreehal-claude-skills
jagreehalLicense: MIT
31

Install this agent skill to your local

pnpm dlx add-skill https://github.com/jagreehal/jagreehal-claude-skills/tree/HEAD/skills/code-flow-analysis

Skill Files

Browse the full folder contents for code-flow-analysis.

Download Skill

Loading file tree…

skills/code-flow-analysis/SKILL.md

Skill Metadata

Name
code-flow-analysis
Description
>-

Code Flow Analysis

Critical rules

  • Trace before implementing non-trivial changes. No abstractions like "the handler" — use file:line.
  • Diagrams stay short (5-15 lines). Show fn(args, deps), Result ok/err, workflow steps, error mark.
  • Confirm understanding with the user before changing code.
  • Red flags: fix without file:line; adding retry/validation/cache without checking upstream; changing internals without callers; assuming Result is ok.
  • Before drawing examples or pattern-specific traces, read references/examples.md.

Workflow

  1. Trace: entry point, fn(args, deps) chain, error/change location, workflow steps, Result propagation — all with file:line.
  2. Draw a concise execution diagram (see examples).
  3. Verify: "Here's the flow: [diagram]. Error/change at [file:line] when [condition]. Correct?" Wait for confirmation.
  4. Only then implement (feeds debugging-methodology / tdd-workflow).

Resources

  • references/examples.md — diagram shape, Result/workflow traces, bug/feature examples, rationalizations. Read before tracing non-trivial flows.

Validation

  • [ ] Entry and full call chain with file:line
  • [ ] Result propagation and error/change location marked
  • [ ] Diagram concise (5-15 lines)
  • [ ] User confirmed understanding before code changes

Constraints

  • Skip for typos, formatting, docs-only.
  • Related: debugging-methodology, tdd-workflow, result-types, fn-args-deps, resilience, validation-boundary, confidence-levels.