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.
git clone --depth 1 https://github.com/jgt87/local-llm-mcpWrote 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/commands/jgt87/local-llm-mcp/local_triage)<a href="https://agentmods.dev/commands/jgt87/local-llm-mcp/local_triage"><img src="https://agentmods.dev/badge/commands/jgt87/local-llm-mcp/local_triage/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/commands/jgt87/local-llm-mcp/local_triage"><img src="https://agentmods.dev/badge/commands/jgt87/local-llm-mcp/local_triage.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 | $0.00020 | $0.00422 |
| Opus 5.5 | $0.00008 | $0.00169 |
| Sonnet 5.5 | $0.00004 | $0.00084 |
| Haiku 4.5 | $0.00002 | $0.00042 |
Grade A, and why
local_triage 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 5d 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
Arguments: $ARGUMENTS
The first argument is the comma-separated label set; anything after it is the source. If the source is a file path, read the file. If it is a shell command, run it and use its output. Otherwise treat it as the text itself. With no source, use the most recent list of items in this conversation (log lines, files, errors). If fewer than two labels are given, ask the user for the label set before calling.
Split the source into items — one per line or entry — and call local_classify once per item with the labels. Independent calls can run in parallel. Report a table of item and label, then list separately the items that came back declined=true (none of the labels fit) and matched=false (unclassified). Do not assign those yourself without saying so. A returned label is triage, not a verdict: spot-check a few, and flag any you disagree with.
Ground rules:
- Use the local-llm tools for the generation step; do the reading, trimming and checking yourself. The local model has no file access, no repo context and no memory between calls, so every call must carry all the text it needs.
- Keep maxTokens tight: locally it runs at roughly 16 tokens per second.
- Stay on the default provider unless the user asked for OpenRouter. Never send secrets, keys or private data to provider="openrouter".
- If a call fails because Ollama is unreachable, say so and stop; do not silently do the work yourself instead.
- Treat the reply as a draft. Check it against the source before presenting it, fix what is wrong, and say briefly what you changed.
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.
- 5d ago First seen · 21 lines · 20 tokens per session scan A aff5a61e1cfe
local_triage is a command published in the GitHub repository jgt87/local-llm-mcp (0 stars, last pushed 6d ago), licensed MIT. It adds 20 tokens to every session and 422 once invoked, about $0.0001 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 2026-10-27.
Other commands, from other repositories
qa-changes
This skill should be used when the user asks to "QA a pull request", "test PR changes", "verify a PR works", "functionally test changes", or when an automated workflow triggers QA validation of code changes. Provides a structured methodology for setting up the environment, exercising changed behavior, and reporting…
react-native
Command "react-native" from getsentry/toolkit, covering xcode build step (react-native xcode) and important notes.
auditar-cascading
Invoca cascading-failures-auditor — audita serviço para 5 triggers de cascade (sem timeout, retry sem jitter, sem circuit breaker, sem deadline, queue unbounded) e gera CASCADING-AUDIT.md.
http-service
Build, review or debug a Bun HTTP service. Loads the http-service skill, then works the task through its workflow.
doctor
Diagnosticar y reparar problemas del framework Don Cheli, git y entorno. Usa cuando el usuario dice "doctor", "problemas del framework", "don cheli no funciona", "repair Don Cheli", "debug setup", "setup broken", "framework broken", "reparar entorno". Detecta y repara issues de configuración, git y dependencias…
doctor
Badi configuration validation. Checks all Badi components and produces a diagnostic report.