Switchyard is a model-routing library and proxy that chooses which language model should handle each LLM request, based on factors such as cost and task suitability. It is for developers running LLM applications who want to route calls through existing gateways, custom harnesses, or a standalone proxy while retaining OpenAI and Anthropic client compatibility. The catalogue add-ons are workflows for operating or integrating Switchyard with coding agents and gateways.
Getting it into your agent
It runs from inside its repository, so the clone comes first — what it calls does not travel with the file alone.
git clone --depth 1 https://github.com/NVIDIA-NeMo/Switchyardnpx agentmods add skills/nvidia-nemo/switchyard/harbor-f2p-p2p-deep-diveWrote 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/nvidia-nemo/switchyard/harbor-f2p-p2p-deep-dive)<a href="https://agentmods.dev/skills/nvidia-nemo/switchyard/harbor-f2p-p2p-deep-dive"><img src="https://agentmods.dev/badge/skills/nvidia-nemo/switchyard/harbor-f2p-p2p-deep-dive/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/nvidia-nemo/switchyard/harbor-f2p-p2p-deep-dive"><img src="https://agentmods.dev/badge/skills/nvidia-nemo/switchyard/harbor-f2p-p2p-deep-dive.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 MCP Rug Pull · line 265 Docker image references without a specific tag (:latest is implicit) or digest (@sha256:...) can be silently replaced by a malicious image.Fix: Pin the image: image:tag or image@sha256:abc123
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 | $0.00238 | $0.04997 |
| Opus 5 | $0.00119 | $0.02499 |
| Sonnet 5 | $0.00048 | $0.00999 |
| Haiku 4.5 | $0.00024 | $0.00500 |
Grade A, and why
harbor-f2p-p2p-deep-dive 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 11d 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 — 285 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Harbor F2P/P2P Deep Dive
Diagnose why an agent failed a Harbor task whose verifier uses SWE-Bench-style F2P/P2P reference tests. Don't assume a failure mode from the score — split failures by class (F2P vs P2P) and trace each through the actual implementation. Produce a root-cause summary per failing test.
Prerequisites
harbor-lab must be available. From this repo: uv run harbor-lab. From outside the repo: harbor-lab if installed on PATH, otherwise ~/path/to/harbor-lab/.venv/bin/harbor-lab. Set a HLAB shell variable to whichever form works (e.g. HLAB="uv run harbor-lab") and use $HLAB throughout.
Test classes: F2P vs P2P
Harbor verifiers (following SWE-Bench convention) typically split reference tests into two classes:
| Class | What it is | What a failure means |
|---|---|---|
| F2P (FAIL_TO_PASS) | Tests that fail on the unpatched repo and must pass after the agent's edit | Agent didn't implement the requested feature/fix (or implemented it incorrectly) |
| P2P (PASS_TO_PASS) | Tests that pass before and must still pass after | Regression — agent broke unrelated existing behavior |
A trial is "resolved" only when all F2P pass AND all P2P still pass. The two failure classes have very different diagnostic paths, so split failures by class before triaging:
- P2P failure → start with regression analysis (the default verdict): the agent touched too much, didn't bound its edits, or its change cascaded. Apply F2P-style validity questions only in narrow cases, all of them rare:
- Hidden fixture or path dependency the test relies on, never mentioned in the instruction (a reasonable refactor breaks the path).
- Pre-existing flaky test (timing, network, unseeded randomness) where the agent's change merely shifted execution.
- Internal-implementation assertion (
assert _private_helper() == "old") rather than a public-contract assertion — agent legitimately refactored the private path. - Pre-broken test that was passing for unrelated reasons and the agent's edit removed the masking.
- Instruction wording that forces a change the P2P test catches — agent had no choice (borderline; depends how forcing the wording is).
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.
- 11d ago First seen · 285 lines · 238 tokens per session scan A 84ab33c43fc2
harbor-f2p-p2p-deep-dive is a skill published in the GitHub repository NVIDIA-NeMo/Switchyard (2,793 stars, last pushed yesterday), licensed Apache-2.0. It adds 238 tokens to every session and 4,997 once invoked, about $0.0012 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 2026-08-30.
Other skills, from other repositories
systematic-debugging
4-phase root cause debugging: understand bugs before fixing.
langsmith-observability
LLM observability platform for tracing, evaluation, and monitoring. Use when debugging LLM applications, evaluating model outputs against datasets, monitoring production systems, or building systematic testing pipelines for AI applications.
experimental-code-coverage-local-debugger
Runs code coverage locally via Universal Test Runner (UTR) or helper scripts, mimicking LUCI trybots. Activate when CQ tryjobs fail or underreport coverage, to test local GN/recipe repairs before uploading, or to debug hermetic crashes.
adversarial-reviewer
Adversarial code review that assumes bugs exist and hunts for them. Use when asked to review code, find bugs, audit for correctness, stress-test a PR, or when someone says "tear this apart" or "what's wrong with this". Give no benefit of the doubt — every line is guilty until proven innocent.
cli-e2e
Write, modify, or debug Docker-based Composio CLI end-to-end tests under ts/e2e-tests/cli, including binary invocation, fixture isolation, output assertions, and package manifests. Use for CLI E2E test suites only; use cli-command for CLI source implementation.
work
Deliver one maintainer-approved EmDash issue, choosing the bug-fix path for a defect and the direct implementation path for an enhancement or task.