Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add skills/muzafferkadir/nextjs-loop-engineering-starter/loop-plannpx skills add muzafferkadir/nextjs-loop-engineering-starter --skill loop-plangit clone --depth 1 https://github.com/muzafferkadir/nextjs-loop-engineering-starterWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/muzafferkadir/nextjs-loop-engineering-starter/loop-plan)<a href="https://agentmods.dev/skills/muzafferkadir/nextjs-loop-engineering-starter/loop-plan"><img src="https://agentmods.dev/badge/skills/muzafferkadir/nextjs-loop-engineering-starter/loop-plan.svg" alt="Measured on agentmods" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00087 | $0.00718 |
| Opus 5 | $0.00044 | $0.00359 |
| Sonnet 5 | $0.00017 | $0.00144 |
| Haiku 4.5 | $0.00009 | $0.00072 |
Grade A, and why
loop-plan scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 4d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 68 lines — stays where its author put it; the contents beside it link to each section on GitHub.
loop-plan — Planner Skill (spec before code)
Turn ONE short idea into a sprint contract: concrete, checkable acceptance criteria in BACKLOG.md. Write no code. Open no PRs.
Workflow
1. Take the idea
- Input: the idea handed to you directly, or the top item under
## Ideasin BACKLOG.md. - If STATE.md contains
loop: paused— stop.
2. Study precedent — has similar work been done?
- BACKLOG-DONE.md — similar completed features, their criteria count and scope
- loop-run-log.md — how many runs/rejects comparable features took
git log --oneline -20— the shape of comparable diffs
A feature similar to one already delivered in a single run is the strongest "fits one run" signal.
3. Size decision — does it fit ONE loop run?
"Fits one run" means all of:
- ≤ 5 acceptance criteria, every one testable
- One area of the app (one route/component cluster + its data)
- No schema change combined with large UI work
- Comparable to a past single-run feature (step 2)
Fits → step 4. Too broad → step 5.
4. Write the spec (fits one run)
- Draft the F-XXX entry with the next free F-number, using the template
in BACKLOG.md: concrete, checkable acceptance criteria; keep the
pnpm snapline for UI features; flag schema changes (human approval per LOOP.md Human Gates). - Add it to BACKLOG.md
## Backlogwith a priority and a precedent note ("similar to F-YYY, delivered in one run"). Remove the idea from## Ideasif it came from there. - Append one line to loop-run-log.md (
— Planner).
5. Split proposal (too broad) — ask the human
- Break the idea into the smallest set of independently shippable features, each passing the step-3 size test, ordered so that earlier features never depend on later ones.
- Do NOT add them to the Backlog yet.
- Get approval:
- Interactive session → present the split (each proposed F-XXX with a one-line scope) and ask directly.
- Scheduled (non-interactive) run → write the proposal under "Waiting on Human" in STATE.md and stop.
- Only after human approval: add the approved features to
## Backlog.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 4d ago First seen · 68 lines · 87 tokens per session scan A 59d69dd1ac02
loop-plan is a skill published in the GitHub repository muzafferkadir/nextjs-loop-engineering-starter (10 stars, last pushed 2mo ago), licensed MIT. It adds 87 tokens to every session and 718 once invoked, about $0.0004 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
deliver
Drive a whole phase or feature-set to "merged + Done" — themed batching over /drive, front-loaded architectural questions, phase-gate ceremonies, and lossless resume. The long-horizon recipe ABOVE /drive; pair with native /goal " delivered" for hands-off, multi-turn completion.
drive
Drive one tracker ticket end-to-end — claim, plan, implement, verify, cross-review, and ship — autonomously, pausing only for architectural decisions. The per-ticket recipe; pair with native /goal " is merged" for unattended multi-turn completion.
auto-review
Drain the orchestrator review-queue — for each task awaiting review, run the code-reviewer, optionally a Codex/Gemini second opinion on CRITICAL.md paths, compose a verdict, and advance it to reviewed or escalate an architectural finding to you. Interactive session only.
autopilot
Take a product idea from the discovery interview all the way to a running delivery loop — drive the spec chain (BRIEF → PRD → SPEC → CONTEXT → phases.yaml → tracker), pausing ONLY at the human approval gates, then roll into /deliver. The single front door for a new forge product; pair with native /goal for hands-off…
second-opinion
Get a second opinion on a plan, diff, or specific files from Claude, Codex, or Gemini (whichever is configured as settings.agents.reviewhostcli, a different host than the primary worker). The reviewer reads the working tree directly, so it can reason about both the proposed change AND surrounding production code…
audit
Fan out N read-only subagents to audit the codebase for safe simplification (dead code, duplication, over-exports, complexity, dependency bloat, stale docs), then assemble a filtered work-order of findings. READ-ONLY — never edits source. The work-order is the deliverable.