Subhajit Pradhan
ResumeEmail
Projects

Developer tools and AI workflow engineer

SmritiFlow

SmritiFlow is a CLI for maintaining living repository memory for coding agents. It scans a codebase, writes structured artifacts, and generates concise agent-facing docs so work can resume with current context instead of guesswork.

Metrics

Proof in numbers

Distribution
npm CLI + Agent Skill
Commands
init, scan, refresh, status, resume
Artifacts
JSON + AGENTS + docs/ai
CLI aliases
smritiflow, sf

Workflow proof

Repository memory loop

SmritiFlow turns repository scans into structured memory and concise agent handoff docs.

  1. Initialize
  2. Scan repo
  3. Refresh changes
  4. Resume work

Architecture

System shape

A TypeScript monorepo ships a Node.js CLI, repo parser, git utilities, core workflow commands, artifact generators, and shared constants. Generated JSON artifacts feed concise docs and agent instructions.

  1. CLI command

    Developers run init, scan, refresh, status, or resume from the target repository.

  2. Repo parser

    The scanner detects project structure, routes, imports, and relevant files.

  3. Git context

    Git utilities identify changed files and freshness signals.

  4. Artifact generator

    Structured cache, project map, scan report, AGENTS.md, and docs/ai files are written.

  5. Agent resume

    Agents read generated artifacts to resume with current context and next actions.

Proof

  • Published as the smritiflow CLI with smritiflow and sf command aliases.
  • Generates cache, project map, scan report, AGENTS.md, and AI handoff docs.
  • Includes init, scan, refresh, status, and resume workflows for agent handoffs.
  • Exposes a repo-hosted Agent Skill through npx skills add subhajitlucky/smritiflow.

Problem

Long-running codebase work often loses context between sessions. Agents and developers need a current, structured view of files, changes, active areas, and next actions instead of rebuilding context from scratch.

Users and context

Built for developers using coding agents who need repeatable repository scans, freshness checks, and handoff summaries.

Product workflow

A developer initializes SmritiFlow in a repo, runs scan or refresh after meaningful changes, checks status before resuming, and uses resume to get focused context for the next work session.

Key engineering decisions

  • Split parsing, git inspection, generators, and command workflows into packages so each layer can be tested independently.
  • Generated both machine-readable JSON and human-readable docs because agents need structure while developers need quick summaries.
  • Added a short sf alias for repeated terminal use without making the primary command ambiguous.
  • Packaged a repo-hosted Agent Skill so compatible agents know the scan, status, refresh, and resume workflow.

Tradeoffs and limitations

  • Generated summaries can become stale, so the CLI includes status and refresh commands instead of pretending context stays fresh forever.
  • Repository scanning is useful for orientation, but it should complement direct code inspection rather than replace it.
  • A CLI-first interface is efficient for developers, but future UI or PR-comment output could improve team adoption.

Next improvements

  • Add richer stale-signal scoring across branches and uncommitted changes.
  • Generate compact PR handoff summaries.
  • Add more language-specific route and dependency extraction.