agentsclimarketplace

Agentctx

Skill rahulkolekardev/agentctx/skills/agentctx

Make your repo agent-ready for Claude Code, Codex, Copilot, Cursor, and AI coding agents.

Install
npx -y skills add rahulkolekardev/agentctx --skill agentctx

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Make a repository ready for AI coding agents by scanning project facts and generating AGENTS.md, CLAUDE.md, GEMINI.md, Copilot instructions, Cursor rules, repo maps, doctor reports, and task-specific context packs. Use when preparing repos for Claude Code, OpenAI Codex, Gemini CLI, Google Antigravity, GitHub Copilot, Cursor, Windsurf, Amp, Aider, Cline, Roo Code, Continue, OpenHands, Replit Agent, Vercel Skills, or other coding agents.

The file declares its own license as MIT. That is the author’s claim about this one file, and it is not the same thing as the license GitHub reports for the repository, which is listed with the other numbers below.

SKILL.md

4.9 KB, as published. Nobody here has run it

AgentCtx Skill

Use this skill when the user asks to make a repository AI-agent-ready, generate AGENTS.md, create CLAUDE.md or GEMINI.md, improve Copilot/Cursor instructions, inspect repository context, create a repo map, check stale agent instructions, or build a task-specific context pack for a coding agent.

Supported agents and outputs

Use the most specific output when available:

  • OpenAI Codex, Amp, Aider, Google Antigravity, Windsurf, Cline, Roo Code, Continue, OpenHands, Replit Agent, Zed Agent Panel, and generic coding agents: AGENTS.md, .repoctx/repo-map.md, and task packs.
  • Claude Code: CLAUDE.md.
  • Gemini CLI and Gemini-based coding workflows: GEMINI.md.
  • GitHub Copilot / Copilot coding agent: .github/copilot-instructions.md.
  • Cursor: .cursor/rules/project.mdc.
  • Web/docs LLM discovery: llms.txt.

When a tool has its own custom-instruction UI but no stable repo-file convention, use AGENTS.md, llms.txt, and .repoctx/packs/*.md as the portable source of truth.

Core principle

Generate minimal, evidence-backed context. Do not create long prompt dumps. Do not invent commands, frameworks, scripts, or architecture details.

Safety rules

  • Never read, print, summarize, or copy .env files.
  • Never include API keys, tokens, private keys, or secrets in generated context.
  • Never guess commands. Use only commands detected from package.json scripts and lockfiles.
  • If a fact is uncertain, mark it as uncertain or omit it.
  • Keep context files short and practical.
  • Prefer deterministic helper scripts when available.

Available helper scripts

From the skill directory, use these scripts when possible:

node scripts/agentctx.mjs scan --root <repo>
node scripts/agentctx.mjs generate --root <repo>
node scripts/agentctx.mjs map --root <repo>
node scripts/agentctx.mjs doctor --root <repo>
node scripts/agentctx.mjs pack "<task>" --root <repo>

Individual scripts are also available:

node scripts/scan-repo.mjs --root <repo>
node scripts/generate-context.mjs --root <repo>
node scripts/doctor.mjs --root <repo>
node scripts/create-pack.mjs "<task>" --root <repo>

Workflow: scan a repo

  1. Run the scanner if scripts are available:

    node scripts/agentctx.mjs scan --root .

  2. Inspect the detected facts:

    • package manager
    • framework
    • language
    • scripts
    • test runner
    • linter
    • database tooling
    • CI
    • important files

  3. Report only facts supported by evidence.

Workflow: generate agent context files

  1. Scan the repo.
  2. Generate only requested files, or defaults:
    • AGENTS.md
    • CLAUDE.md
    • GEMINI.md
    • .github/copilot-instructions.md
    • .cursor/rules/project.mdc
  3. Include:
    • project overview
    • detected commands
    • verification checklist
    • important files
    • safety rules
  4. Do not overwrite files unless user explicitly asks.

Preferred command:

node scripts/agentctx.mjs generate --root .

Workflow: doctor report

Use this when the user asks whether the repo is ready for AI agents:

node scripts/agentctx.mjs doctor --root .

Check:

  • missing context files
  • stale commands
  • package-manager contradictions
  • oversized context
  • secret-like content
  • missing test/build/lint scripts
  • missing README, license, or CI

Workflow: task context pack

Use this when the user asks for a context pack for a specific task:

node scripts/agentctx.mjs pack "add login with Google OAuth" --root .

The pack should include:

  • task summary
  • repository summary
  • likely relevant files with reasons
  • relevant commands
  • detected patterns
  • safety notes
  • verification checklist

Do not dump entire files unless explicitly requested.

Reference files

Use supporting references only when needed:

  • references/detection-rules.md
  • references/security-rules.md
  • references/templates.md
  • references/output-formats.md

Response style

When finished, summarize:

  1. What was created or checked.
  2. Key detected facts.
  3. Any warnings or missing items.
  4. Exact next commands the user can run.

Keep looking

Skills are one crate of 328,083. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.