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 skills add restarter/lets-workflow --skill loop-testgit clone --depth 1 https://github.com/restarter/lets-workflowWrote 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/restarter/lets-workflow/loop-test)<a href="https://agentmods.dev/skills/restarter/lets-workflow/loop-test"><img src="https://agentmods.dev/badge/skills/restarter/lets-workflow/loop-test/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/restarter/lets-workflow/loop-test"><img src="https://agentmods.dev/badge/skills/restarter/lets-workflow/loop-test.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector warn
SkillSpector: 1 finding, up to medium
These are SkillSpector’s own severities. On a checked sample its high-severity flags on skills were ~96% false positives — a documented command, a public API, a “never do X” rule — so we show them as a caution to read, not a verdict. Why →
- medium Excessive Agency · line 134 Skill allows unbounded resource consumption (API calls, storage, compute). Without rate limits or quotas, a compromised or misbehaving agent can cause denial-of-service or cost overruns.Fix: Set explicit rate limits, timeouts, and resource quotas for API calls, file operations, and compute. Implement circuit breakers for runaway loops.
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 5.1 | $0.00112 | $0.02303 |
| Opus 5.5 | $0.00045 | $0.00921 |
| Sonnet 5.5 | $0.00022 | $0.00461 |
| Haiku 4.5 | $0.00011 | $0.00230 |
Grade A, and why
loop-test 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 1mo 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 — 173 lines — stays where its author put it; the contents beside it link to each section on GitHub.
loop-test (local experiment)
/loop schedules periodic re-invocations of the same prompt. Two modes:
| Mode | Syntax | Mechanic |
|---|---|---|
| Interval | /loop 5m /foo · /loop /standup 1h · /loop check the deploy every 20m |
Cron entry in .claude/scheduled_tasks.json. Default 10m when only [interval] shorthand. Survives session restart (durable). |
| Dynamic / self-paced | /loop check the deploy (no interval anywhere) |
Model calls ScheduleWakeup(delaySeconds, prompt, reason) at the end of each turn. Skip → loop ends silently. |
/loop is the wrong tool for sequential workflow phases. It repeats the same prompt each tick — to chain plan → review → check, the prompt itself has to be phase-aware (read state, advance, write back). For state-convergent autonomy, use /goal instead (see goal-test skill).
Right tool for /loop:
- Polling external state (CI runs, deploys, PR comments)
- Periodic checks (every 20m: any new bd-ready tasks?)
- Babysitting a long-running job (check progress, surface failures)
- Continuous monitoring during a long focused session
How /loop actually works
Reverse-engineered from claude binary v2.1.150.
Dynamic-mode contract
This is the mode you care about most — interval is just cron.
- User runs
/loop <prompt>(no interval). - Model does the work.
- Before yielding, model calls
ScheduleWakeup(delaySeconds, prompt, reason):delaySecondsclamped to[60, 3600]promptmust be exactly the same string as the original/loopinputreasonis one sentence shown to user + sent to telemetry — be specific ("polling CI run"beats"waiting")
- Harness re-invokes the model at the scheduled time. Loop continues.
- Skip
ScheduleWakeup→ loop ends silently. No error.
Cache TTL gotcha
Anthropic prompt cache TTL is 5 minutes. Picking delaySeconds:
< 270s→ cache stays warm, cheap re-invocation>= 1200s→ commit to the miss, one cache-miss amortizes over a long wait300sis worst-of-both — you pay the miss without amortizing. Avoid exactly 300.
What ships with it
1 file beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
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.
- 1mo ago First seen · 173 lines · 112 tokens per session scan A 0525b5d511aa
loop-test is a skill published in the GitHub repository restarter/lets-workflow (17 stars, last pushed today), licensed MIT. It adds 112 tokens to every session and 2,303 once invoked, about $0.0004 per session on Opus 5.5. 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 2026-08-30.
Other skills, from other repositories
review-work
Post-implementation gate review: run manual QA on the real surface yourself, then launch ONE gate reviewer (never a panel) to audit goal, constraints, code quality, security, missed context, and QA evidence. Use before a PR handoff or when the user explicitly asks to review completed work.
brooks-sweep
Full-sweep mode: runs a unified analysis across all quality dimensions — code decay, architecture, tech debt, and test quality — then applies fixes directly to the codebase. Safe changes are auto-applied; risky changes are confirmed before execution. Drawing on twelve classic engineering books. Triggers when: user…
brooks-test
Test quality review drawing on twelve classic engineering books — with primary focus on xUnit Test Patterns, The Art of Unit Testing, How Google Tests Software, and Working Effectively with Legacy Code — that diagnoses structural problems in an existing test suite: brittleness, mock abuse, coverage illusions, slow…
include-test-files-that-assert-on-behavior-being-changed-in-decl
When delegating a task affected by this skill, include.
spike-consumer-adversarial
OI-3 spike harness — heavy consumer, ADVERSARIAL arm. Worst-case early-exit test: the mid-workflow Skill call has no continuation guardrail and the guidance skill ends with a final-sounding anchor. Use only when explicitly invoked by the spike harness with a TRIALID and data path.
crap-analyzer
Use to produce a risk-based refactor + test plan for recently-changed code on a diff/branch/PR by computing CRAP (complexity × untested) on changed methods. Multi-language — TypeScript, JavaScript, Python, Java, Kotlin, Go, Ruby, C#, Rust, PHP — auto-discovers how the repo generates coverage. Triggers …