AgentStack
Browse Sign in
Browse Why AgentStack Sell Docs
Sign in
SKILL verified MIT Self-run

Coding Backend Cursor

skill-ahmadrosid-nakama-coding-backend-cursor · by ahmadrosid

Runtime prompt layer for Cursor Agent CLI coding agent runs.

— No reviews yet
0 installs
0 views
— view→install

Install

$ agentstack add skill-ahmadrosid-nakama-coding-backend-cursor

✓ scanned · ✓ verified, works with Claude Code, Cursor, and more.

Security review

✓ Passed

No issues found. Passed automated security review. · v0.1.0 How review works →

  • ✓ Prompt-injection patterns
  • ✓ Secret / credential exfiltration
  • ✓ Dangerous shell & filesystem operations
  • ✓ Untrusted network calls
  • ✓ Known-malicious package signatures

What it can access

  • ✓ Network access No
  • ✓ Filesystem access No
  • ✓ Shell / process execution No
  • ✓ Environment & secrets No
  • ✓ Dynamic code execution No

From automated source analysis of v0.1.0. “Used” means the capability is present in the source — more access means more to trust, not that it’s unsafe.

View the full security report →

Verified badge

Passed review? Show it. Paste this badge into your README, it links to the public security report.

AgentStack Verified badge Links to your public security report.
[![AgentStack Verified](https://agentstack.voostack.com/badges/verified.svg)](https://agentstack.voostack.com/security/report/skill-ahmadrosid-nakama-coding-backend-cursor)

Reliability & compatibility

✓ Security review passed
0 installs to date
— no reviews yet
● today

Declared compatibility

Claude CodeClaude Desktop

Compatibility is declared by the source manifest. End-to-end runtime verification is coming, see below.

Preview Execution monitoring

We're building live execution health for every listing: tool-call success rate, median latency, uptime, and last-checked timestamps, measured, not self-reported. It isn't live yet, so we don't show numbers we can't stand behind.

How agent discovery & health will work →
Are you the author of Coding Backend Cursor? Claim this listing to set pricing, connect Stripe payouts, and keep 70% of every sale.
Sign up to claim

About

You are preparing a coding agent run for Cursor Agent CLI (agent), orchestrated via the bash tool.

Prerequisites

  • Cursor Agent CLI must already be installed and authenticated on the Nakama server host.
  • Verify with: agent --version
  • Nakama does not auto-install Cursor Agent and does not inject Nakama provider credentials. Host Cursor auth is required.
  • If agent is missing or unauthenticated, tell the user to install and authenticate Cursor Agent CLI themselves, then retry. Do not run npm install -g for this backend.

Repo setup (do this before coding)

Bash defaults to the profile workspace. Cursor Agent must work on a real git checkout there — not on soul files alone.

  1. Identify the target repo (URL or folder name from the user). If unclear, ask once.
  2. Check whether that repo already exists under the workspace (ls, or test -d /.git).
  3. If it is missing, clone it into the workspace (git clone ), then continue.
  4. Hand off soon: give Cursor a short brief (issue URL + goal + constraints). Do not fully re-solve the bug with file tools first.

Do not invent a repo URL. Do not run coding work against an empty workspace when the user named a remote repo.

Command (required shape)

Set bash cwd to the repo directory. Keep argv0 as agent — never cd … && agent (with codingAgent: true, Nakama requires the harness binary first).

agent -p 'Implement the requested change and summarize what you verified' --output-format text --yolo

bash args:

  • cwd: absolute path to the repo checkout inside the profile workspace
  • codingAgent: true
  • timeoutMs: often 600000–1800000 for substantial runs

Flags:

  • -p / --print — non-interactive one-shot
  • --output-format text — short final answer for Nakama (preferred). stream-json also works: Nakama summarizes the NDJSON into assistant/tools/result and saves the full log under artifacts/coding-agent-runs/.
  • --yolo — required for unattended background dispatch

Commits and pull requests

Cursor Agent can create branches, commit, push, and open GitHub PRs with git + gh on the host (same capabilities as a local checkout). For implementation runs (feature / fix / repo change), include that in the -p brief by default — do not stop at uncommitted edits on main. Skip commit/push/PR only when the user explicitly asked for local-only work (exploration, draft, "just look", "don't push").

For implementation runs, put this in the agent prompt (adapt issue number, title, and scope):

agent -p 'Fix issue #175: .

When done:
1. Create a feature branch off the default branch (do not commit on main).
2. Commit with a clear message focused on why.
3. Push the branch and open a PR with gh (link Fixes #175 in the body).
4. Reply with the PR URL, what changed, and what you verified.

Do not force-push. Do not skip hooks.' --output-format text --yolo

Rules:

  • Require commit/push/PR by default for implementation runs. Skip git publish steps only when the user explicitly asked for local-only work (exploration, draft, "just look", "don't push").
  • Prefer one coding-agent run that implements and opens the PR. If the first run left uncommitted work, follow up with another agent -p (or bash gh/git) in the same repo cwd to finish ship — verify with git status / gh pr view first.
  • Host needs gh authenticated for the target remote. If push/PR fails on auth, tell the user to fix host gh auth / git credentials, then retry.
  • Never invent a PR URL. Confirm from agent stdout or gh pr list / gh pr view.

After the run

  • Summarize the returned stdout for the user (already summarized if stream-json was used).
  • If the result is unclear, verify completion with git status / git diff --stat in the repo (via bash cwd), then report what changed. Use the full log path from stdout when you need raw events.
  • If a PR was requested: confirm branch is pushed and report the PR URL (from stdout or gh pr view --json url -q .url). If changes exist but no PR, finish shipping — do not claim the issue is done.
  • If the run failed (non-zero exit, timeout, auth error, or empty useful output), explain clearly and ask the user to fix host install/auth when that is the cause.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

Install and usage instructions live in the source repository linked above.

Reviews

No reviews yet, be the first.

Versions

  • v0.1.0 Imported from the upstream source.