CategoryWeightCategory scoreContribution

Securityprompt, exec, net, exfil, eval

35%

30.8 pts

Supply chainhash, typosquat, maintainer, lockfile

20%

100

20.0 pts

Maintenancestaleness, pinning, CI

15%

100

15.0 pts

TransparencySKILL.md, perms, README

15%

100

15.0 pts

Communityinstalls, verify, response

15%

100

15.0 pts

Findings & checks · 1 flagged

Securityscore 88 · 1 finding

MEDIUMInstruction telling the agent not to ask for approvalSS-SKILL-INJECT-DONT-ASK-01 · Prompt injection · llms-install.md

MEDIUM — it fires on intent; the real damage depends on the host agent's own approval-gating.

Why it matters

The text {match} tells the agent to skip the normal "ask the user first" gate. Used adversarially it removes the human-in-the-loop check before destructive or sensitive actions, turning a normally-gated agent into a fire-and-forget executor.

The exact value spotted

excerptllms-install.md· markdown

27- Start with usage visibility, AI health checks, and cost-saving suggestions.

28- Do not invent usage, billing, savings, or provider numbers.

29- Do not ask for provider API keys first; introduce BYOK only after value is established or

… (30 chars elided on L29)

30- Runtime forwards to `@skillfm/mcp@latest` with `SKILLFM_GATEWAY_SOURCE=llm-cost-optimizer`

… (1 chars elided on L30)

Occurrences

1 occurrence · at L29

How to fix

Remove the approval-skipping instruction, or scope it narrowly to a specific safe, reversible action.

Delete blanket "don't ask / no need to confirm" directives from the skill.
If the skill is a genuine autonomous job, restrict the opt-out to a named non-destructive action rather than all actions.

Framework references

OWASPLLM01 ATLASAML.T0051

Trace & refs

ruleSS-SKILL-INJECT-DONT-ASK-01sha25668b030f55bcd484a…rubric 365aacaView on GitHub

Supply chainscore 100 · 0 findings

✓—All supply chain checks passedNo findings in this category for the latest scan.pass

Maintenancescore 100 · 0 findings

✓—All maintenance checks passedNo findings in this category for the latest scan.pass

Transparencyscore 100 · 0 findings

✓—All transparency checks passedNo findings in this category for the latest scan.pass

Communityscore 100 · 0 findings

✓—All community checks passedNo findings in this category for the latest scan.pass

The primary manifest — the file an agent reads to learn what this artifact does.

README.md · 1 file

README.md3.2 KB · Markdown

SkillFM LLM Cost Optimizer MCP

LLM cost optimization and provider usage visibility for MCP-capable agents.

This repository is a transparent SkillFM gateway package. It exists to match a specific marketplace/search intent, then delegates execution to the canonical SkillFM Beacon MCP and local runtime.

Quick Start

Use this gateway as an MCP stdio server:

npx -y @skillfm/llm-cost-optimizer-mcp@latest

Equivalent explicit SkillFM entrypoint:

npx -y @skillfm/mcp@latest

For MCP clients that accept JSON config:

{
  "mcpServers": {
    "llm-cost-optimizer": {
      "command": "npx",
      "args": ["-y", "@skillfm/llm-cost-optimizer-mcp@latest"]
    }
  }
}

What This Gateway Helps With

Turn LLM spend questions into SkillFM Beacon usage and health checks.
Guide users from rough token visibility to provider-level BYOK monitoring when they are ready.
Surface cost-saving opportunities in context, model choice, cache use, batching, and repeated work.
Keep the public package thin while delegating execution to the SkillFM local runtime.

This gateway is for users who ask:

"How do I reduce OpenAI or Anthropic spend?"
"Which agent workflow is burning tokens?"
"Can I optimize LLM cost without sending keys to a SaaS dashboard?"
"How do I connect provider usage after I trust the local checkup?"

How It Works

The public package is intentionally thin. It sets SKILLFM_GATEWAY_SOURCE=llm-cost-optimizer and forwards execution to @skillfm/mcp@latest, which then delegates to the canonical @skillfm/local runtime.

No private SkillFM platform logic, customer data, provider credentials, or backend modules live in this repository.

Activation Path

Start the local sidecar when the user wants to connect SkillFM:

npx -y @skillfm/llm-cost-optimizer-mcp@latest start

Agent-driven activation flow:

Read the sidecar URL from stdout or ~/.skillfm/local.json.
Call POST <sidecar_url>/activate/start.
Show the returned verification URL and real device code to the user.
Poll POST <sidecar_url>/activate/poll until it returns activated: true.
Use SkillFM Beacon MCP tools for checkups, usage visibility, BYOK guidance, and cleanup audits.

Do not fabricate a placeholder code. The device code must come from POST /activate/start.

Security Model

Transparent gateway: This package is a public search/distribution gateway, not a separate hidden product.
Local-first runtime: Runtime execution goes through SkillFM's local package.
BYOK boundary: Provider keys are introduced only after user value is established, and key values are not read back by agents.
No private core: This repository contains distribution metadata, docs, and a thin launcher only.
Read-only first: Cleanup and usage inspection begin with read-only audits.

See PRIVACY.md and SECURITY.md.

Marketplace Keywords

LLM cost optimizer, OpenAI usage, Anthropic usage, token usage, AI cost optimizer, BYOK, SkillFM Beacon, MCP server.

Expected Official MCP Registry single-token hits:

llm-cost
cost
optimizer
openai
anthropic
usage
skillfm

License

MIT

Skillfm Llm Cost Optimizer Mcp— agent skill

Is Skillfm Llm Cost Optimizer Mcp safe to install?

SkillFM LLM Cost Optimizer MCP

Quick Start

What This Gateway Helps With

How It Works

Activation Path

Security Model

Marketplace Keywords

License

Audit the pieces. Scan the whole. Decide.