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/blake-simpson/belmont/loopnpx skills add blake-simpson/belmont --skill loopgit clone --depth 1 https://github.com/blake-simpson/belmontWhat 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 | $0.00047 | $0.13353 |
| Opus 5 | $0.00023 | $0.06677 |
| Sonnet 5 | $0.00009 | $0.02671 |
| Haiku 4.5 | $0.00005 | $0.01335 |
Grade A, and why
loop 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 2d 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 — 700 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Belmont: Loop
Claude Code or Codex only. This skill drives one Belmont feature to completion by repeatedly running the implement → verify → next → status cycle, pausing between iterations so you can watch progress and steer. It is a thin orchestration wrapper around the host tool's long-running interactive primitive:
- Claude Code delegates to the built-in
/loopskill. - Codex delegates to Goal mode with
/goal.
Other AI CLIs do not have an equivalent interactive loop primitive, so Belmont keeps this skill hidden from their default install surface.
If you are neither Claude Code nor Codex, stop and tell the user to run belmont auto --feature <feature> from the terminal instead.
This is the interactive, in-session counterpart to the headless belmont auto CLI (also aliased belmont loop). Use /belmont:loop in Claude Code or $belmont:loop in Codex when you want to stay in the REPL and have the agent advance the feature milestone-by-milestone without you re-typing each skill. Use belmont auto when you want fully headless, parallel, worktree-based execution from the terminal.
Loop is the steering tool, not the throughput tool. Auto will always be faster — it runs milestones in parallel worktrees and gives every phase a fresh context. Loop's value is that you are present and can redirect with a sentence. The recipe below therefore optimises for not wasting your session — batching follow-up fixes, triaging polish out of the critical path, and scoping re-verification — rather than for raw parallelism. Do not try to recover auto's parallelism here.
Milestone structure is immutable outside /belmont:tech-plan
You MUST NOT add, remove, rename, re-scope, or re-parent any ### M<N>: milestone heading in PROGRESS.md. Only /belmont:tech-plan may restructure milestones. Every other skill — implement, verify, next, debug-auto, debug-manual, the triage phase — may only edit tasks inside existing milestone headings.
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.
- 2d ago First seen · 700 lines · 47 tokens per session scan A a96346202d21
loop is a skill published in the GitHub repository blake-simpson/belmont (4 stars, last pushed 14d ago), licensed Apache-2.0. It adds 47 tokens to every session and 13,353 once invoked, about $0.0002 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
agent-performance-benchmarker
Agent skill for performance-benchmarker - invoke with $agent-performance-benchmarker.
agent-quorum-manager
Agent skill for quorum-manager - invoke with $agent-quorum-manager.
agent-agent
Agent skill for agent - invoke with $agent-agent.
agent-crdt-synchronizer
Agent skill for crdt-synchronizer - invoke with $agent-crdt-synchronizer.
agent-adaptive-coordinator
Agent skill for adaptive-coordinator - invoke with $agent-adaptive-coordinator.
agent-benchmark-suite
Agent skill for benchmark-suite - invoke with $agent-benchmark-suite.