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 KempnerInstitute/hpc-agentic-recipes --skill local-searchgit clone --depth 1 https://github.com/KempnerInstitute/hpc-agentic-recipesWrote 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/kempnerinstitute/hpc-agentic-recipes/local-search)<a href="https://agentmods.dev/skills/kempnerinstitute/hpc-agentic-recipes/local-search"><img src="https://agentmods.dev/badge/skills/kempnerinstitute/hpc-agentic-recipes/local-search/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/kempnerinstitute/hpc-agentic-recipes/local-search"><img src="https://agentmods.dev/badge/skills/kempnerinstitute/hpc-agentic-recipes/local-search.svg" alt="Reviewed on agentmods" width="80" 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 5.1 | $0.00073 | $0.00436 |
| Opus 5 | $0.00036 | $0.00218 |
| Sonnet 5 | $0.00015 | $0.00087 |
| Haiku 4.5 | $0.00007 | $0.00044 |
Grade A, and why
local-search 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.
What it actually says
Local search
The built-in WebSearch tool runs on Anthropic infrastructure and does not work against a local endpoint.
vLLM rejects it with 400 ... tools.0.input_schema Field required. SGLang returns 200 and drops the tool,
so no search happens and nothing says so. Either way, use search.sh through Bash instead. It needs no API
key.
Call it as search.sh if it is on PATH, otherwise as bash <repo>/common/tools/search.sh.
Commands
| Command | Use for |
|---|---|
search.sh web "<query>" [n] |
general web search |
search.sh arxiv "<query>" [n] |
preprints, with abstracts |
search.sh crossref "<query>" [n] |
published papers by DOI |
search.sh pubmed "<query>" [n] |
biomedical literature |
search.sh openalex "<query>" [n] |
papers with citation counts |
search.sh wiki "<query>" [n] |
background definitions |
search.sh fetch "<url>" |
plain text of one page |
n is the result count and defaults to 5.
How to use it
- Start with
webfor general questions, orarxivandopenalexfor research literature. - Read the returned titles, URLs, and snippets.
- Run
fetchon the URLs worth reading in full before answering. - Cite the URLs you used.
Run two or three different modes for a literature question, since each source indexes different material. Quote the query; results are plain text, one numbered block per hit.
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 · 38 lines · 73 tokens per session scan A ecf3d62902d1
local-search is a skill published in the GitHub repository KempnerInstitute/hpc-agentic-recipes (11 stars, last pushed 15d ago), licensed MIT. It adds 73 tokens to every session and 436 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
hns-lsel-curator
Local Self-Evolution Loop (LSEL) curator — the CLUSTER + drain engine for the GOOS-local PROPOSE→APPLY seam closure (SPEC-LSEL-LOCAL-EVOLUTION-001). Companion-offset drain of .moai/lessons-inbox.jsonl with a drain-side severity filter that drops the 65% Bash-timeout/sandbox noise, eventkey clustering with a frequency…
moai-workflow-worktree
Git worktree management for parallel SPEC development with isolated workspaces, automatic branch registration, and seamless MoAI-ADK integration. Use when setting up parallel development environments.
hns-workflow-ci-loop
Unified CI watch + auto-fix loop skill. Polls gh pr checks after /moai sync PR creation, classifies required vs auxiliary failures, attempts safe automated patches (max 3 iterations), and escalates semantic failures to the user. Use for CI loop workflow — NOT for general loop iteration patterns (see…
hns-moaiadk-patterns
Skill "hns-moaiadk-patterns" from modu-ai/moai-adk, covering moai-adk-go domain patterns, architecture quick reference, key source paths, pipeline specialist delegation map and template-first build cycle.
moai-harness-learner
Harness learning subsystem coordinator. Produces Tier 4 auto-update proposal payloads consumed by the orchestrator (which surfaces them via AskUserQuestion) and orchestrates Apply/Rollback flows. Triggers when harness learning proposals are pending or learning lifecycle management is needed.
moai-kanban-foreman
One unattended kanban foreman iteration: watch the backlog queue, dispatch the next operator-picked card to an isolated worker, collect completion evidence on read (not on claims), and report. This is the body the project's loop.md driver invokes each iteration of a bare /loop; it can also be invoked directly to test…