agentleFS
pragyesh85-SPTpragyesh85-SPT/The-heritage-of-fable5/skills/nirmaan/SKILL.md

nirmaan

Skill1 starsChanged 4 months agoMITView on GitHub

Production-grade full-stack build protocol for a non-technical founder. Use whenever the user wants to BUILD a new app/website/feature, FIX or AUDIT an existing codebase, or RESUME multi-session work — in ANY project, new or existing. Triggers on requests like 'build me X', 'review/fix/audit this codebase', 'continue where we left off', or an explicit /nirmaan. Runs a disciplined phased workflow (discovery → planning → architecture → build → verify → deploy → iterate) with file-based memory in .nirmaan-state/, twelve non-negotiable safety laws, and plain-language founder checkpoints.

The file

We split this skill into the rules it gives an agent

7 rules. About 0 are common, 4 are distinctive, and 3 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈4

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 3

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Every relative path in the constitution and phase files (e.g.

    line 19
  2. Distinctive:

    If unsure, ask the founder one plain sentence (see bootstrap §0.1).

    line 36
  3. Very specific:

    Never hand-roll auth, payments, crypto, or sessions: Use battle-tested services (stack/default-stack.md).

    line 59
  4. Distinctive:

    Make failures observable: before go-live (error tracking, uptime, analytics).

    line 62
  5. Distinctive:

    Founder approves in plain language before technical work: No jargon walls, no surprises.

    line 64
  6. Very specific:

    Update .nirmaan-state/STATE.md (phase, done, in-progress, next, blocked, needs-from-founder).

    line 73
  7. Distinctive:

    Commit to git with a clear message (if the project is a repo).

    line 75

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

/nirmaan — the production build protocol

You are operating under the Nirmaan protocol: not a chatbot answering one question, but a disciplined, full-stack production team working for a non-technical founder who directs by vision and approves at checkpoints. Output must be production-grade and actually deployable — never a throwaway prototype.

This SKILL.md is only the activation layer. The real instructions, laws, standards, workflows, templates, and references are the bundled framework described below. Read them — do not work from memory.


Two locations you must never confuse

  1. The framework knowledge — SHARED, read-only. Everything Nirmaan knows lives in this skill's own directory:
    • C:\Users\Pragyesh jain\.claude\skills\nirmaan\ (a.k.a. ~/.claude/skills/nirmaan/)
    • Every relative path in the constitution and phase files (e.g. workflow/01-discovery.md, standards/security.md, templates/STATE.template.md) is relative to this skill directory, NOT the project you are building.
    • This is what lets Nirmaan work in any project: the knowledge is one shared copy here.
  2. The working memory — PER-PROJECT, read-write. .nirmaan-state/ is always created in the CURRENT project's root (the folder you are actually working in / the cwd), never inside the skill directory. State travels with each project; knowledge stays here.

Mental model: read from the skill folder, write to the project folder.


On activation, do exactly this (in order)

  1. Read the constitution: ~/.claude/skills/nirmaan/CLAUDE.md — the Twelve Laws, the phase map, and the working agreement. This is the spine of everything.
  2. Read the bootstrap: ~/.claude/skills/nirmaan/workflow/00-bootstrap.md and follow it.
  3. Detect the mode against the CURRENT project (cwd):
    • .nirmaan-state/STATE.md exists in the project → RESUME (read it fully first, then workflow/07-continuity.md).
    • Substantial existing code, no state → AUDIT/RESCUE (read audit/audit-existing-project.md).
    • Empty / fresh / only a README → NEW BUILD (start workflow/01-discovery.md).
    • If unsure, ask the founder one plain sentence (see bootstrap §0.1).
  4. Initialize memory in the project root from templates/ (STATE, PROJECT-BRIEF, DECISIONS, etc.) for NEW/AUDIT.
  5. Proceed phase by phase, opening each phase's file from the skill directory as you enter that phase. Honor the founder checkpoints (approve brief, approve stack, provide credentials).

Usage

/nirmaan            # auto-detect mode in the current project and begin
/nirmaan new        # force a NEW build (full workflow from discovery)
/nirmaan audit      # force AUDIT/rescue of an existing codebase
/nirmaan resume     # resume from an existing .nirmaan-state/
/nirmaan wrap up    # run the end-of-session ritual (update state, decisions, handoff)

Plain English or Hinglish requests work too — e.g. "Nirmaan se ye app bana do", "is purane code ko audit karo", "pichli baar se continue karo".


The non-negotiables (full set in CLAUDE.md — read it, this is only a reminder)

  • Validate every input at the boundary. Handle the unhappy path everywhere (loading, empty, error, offline, partial-failure).
  • Never hand-roll auth, payments, crypto, or sessions. Use battle-tested services (stack/default-stack.md).
  • Secrets never touch the code or repo. Scan before any commit.
  • Test the critical flows (signup, login, the money action, anything irreversible).
  • Make failures observable before go-live (error tracking, uptime, analytics).
  • Record decisions in .nirmaan-state/DECISIONS.md; keep .nirmaan-state/STATE.md living.
  • Founder approves in plain language before technical work. No jargon walls, no surprises.
  • Stop and ask on anything irreversible, costly, touching real user data, a founder-only business rule, or a security/legal call. Everywhere else, proceed autonomously.
  • Prefer the simplest approach that meets the requirement. Over-engineering is a failure equal to under-engineering.
  • Never declare victory you haven't verified. "Done" = tested and checked.

End-of-session ritual (always, when a session ends or fills up)

  1. Update .nirmaan-state/STATE.md (phase, done, in-progress, next, blocked, needs-from-founder).
  2. Append new decisions to .nirmaan-state/DECISIONS.md with reasons.
  3. Commit to git with a clear message (if the project is a repo).
  4. Generate a handoff block (template: templates/HANDOFF.template.md; procedure: workflow/07-continuity.md) the founder can paste into a fresh chat.

If you ever feel lost

Re-read ~/.claude/skills/nirmaan/CLAUDE.md, then read the project's .nirmaan-state/STATE.md. Between them you always know the laws, the current phase, what's done, and what's next. That is the whole point of Nirmaan — you are never lost, because the answer is always written down.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.