Navjyot Nishant

Route application AI jobs through the CLI subscriptions users already pay for, instead of a second token-billed API path.

• Jobs run through authenticated local CLI sessions.
• Outbound bridge polling; no inbound ports or tunnels.
• Claude, Codex, Cursor and Gemini backends behind one `/v1` API.

Measure AI-assisted engineering from commit evidence, using git trailers instead of developer surveillance.

• Commit-level `AI-Attribution` trailer with versioned semantics.
• Local transcript ingest; no prompts, keystrokes or file contents read.
• Homebrew, Scoop, `go install`, local reports and optional DevLake sync.

Reusable Claude Code skills and agents for SDLC work, with human-controlled commits as the design rule.

• Review skills advise only; authoring skills write one artifact.
• No skill runs `git add`, commit, push or tag.
• Outputs trace to Keep a Changelog, INVEST, Scrum, SAFe and OpenSSF OSPS.

Turn account knowledge into live sales planning: contacts, notes, org structure, AI insight and generated account-plan decks.

• PPTX account plans generated from live account data.
• Interactive org chart with bulk manager reassignment.
• RBAC, per-menu permissions, TOTP MFA and Vault-backed secrets.

Claude Cowork plugins for enterprise pursuit and writing teams, packaged as repeatable workflows rather than one-off prompts.

• Marketplace sync keeps installed Cowork plugins current.
• RFP pursuit toolkit for clarification questions and proposal scoring.
• Writing desk with research, fact-check, style and social-post workflows.

An end-to-end agentic development system: from a raw idea or an existing repo to reviewed pull requests, built by a team of role-based agents under enforced guardrails.

• Greenfield from an idea, or brownfield by adopting it inside an existing repo.
• Role-based agents (architect, developers, QA) work up to four tickets at once.
• Breaker, gate and shield hooks enforce guardrails; release promotion stays human.

Cross-session memory for Claude Code: decisions, tasks and learnings carried into the next session, as a plugin rather than a service.

• Plain Markdown store in the repo; no database, server or cloud.
• A session-start hook injects a bounded, provenance-labelled slice of context.
• Writes on meaningful change, never through an LLM call, and fails open on any error.