mcp@4.11.0
VERIFIEDPublisher claims: "Crypto social intelligence from X and Telegram, plus the Elfa Auto condition engine" Observed in sandbox: calls api.elfa.ai only, touches nothing outside the project, writes no memory.
Coverage
Started with a decoy ELFA_API_KEY; tool list observed. Tool calls with a real credential not tested.- Tested
- server startup with a tagged decoy in ELFA_API_KEY · MCP initialize and tools/list · every tool description, for hidden instructions · where the decoy credential went · every file read, file written and host contacted by what ran
- Not tested
- what the tools do when called with a real credential
- Needs
- ELFA_API_KEY — give it to the server where it runs, through vet-skills run --secrets, so the value never enters the server's environment. Never paste a credential into a chat, a ticket or a report.
Before you give it your token
- It will ask for
ELFA_API_KEYELFA_MCP_AUTHELFA_OAUTH_INTROSPECTION_TOKENELFA_OAUTH_INTROSPECTION_URLELFA_OAUTH_ISSUERELFA_OAUTH_SCOPESConfiguration, not credentials: ELFA_BASE_URL, ELFA_EXTRA_HEADERS, ELFA_MCP_ALLOWED_HOSTS, ELFA_MCP_ALLOWED_ORIGINS, ELFA_MCP_HOST, ELFA_MCP_MAX_RESPONSE_CHARS, ELFA_MCP_OPENAI_APPS_CHALLENGE, ELFA_MCP_PORT, ELFA_MCP_RESOURCE_URL, ELFA_MCP_TRANSPORT, ELFA_RETRIES, ELFA_TIMEOUT- Where it talks to
- elfa.ai (its own)
- What the honeypot saw
- no credential files read · reached only api.elfa.ai (its own)
- What we could not test
- what the tools do when called with a real credential
Connect it with a scoped credential, under the guard. It asks for ELFA_API_KEY, ELFA_MCP_AUTH, ELFA_OAUTH_INTROSPECTION_TOKEN, ELFA_OAUTH_INTROSPECTION_URL, ELFA_OAUTH_ISSUER, ELFA_OAUTH_SCOPES and talks to elfa.ai. Give the credential to the guard, not the server: the value never enters the server's environment and only travels to the host you bind.
Never paste a credential into a chat, a ticket or a report, including ours. How the guard holds it.
What changed since 4.10.0
vet-skills diff mcp 4.10.0 4.11.0This version reaches for something 4.10.0 did not: 2 new tools (calendar, prediction_markets) · 4 rewritten tool descriptors (auto_draft, auto_draft_write, auto_query, auto_query_write). Compared from the two recorded runs, not from source.
- New tools
- calendar, prediction_markets
- Rewritten tool descriptors
- auto_draft, auto_draft_write, auto_query, auto_query_write
The guard can hold an update like this until someone approves it (runtime.on_change: hold). A change that only shows with a real credential, or on a path the run did not reach, is not in a trace and so not in this comparison.
Is [email protected] safe to install?
KillerSkills ran it in the KillerSkills honeypot sandbox (2 scenarios) on 2026-11-17 and scored it 93/100 (Verified). It did what it claims and nothing else we could detect: no undeclared network egress, no credential reads, no persistent instruction writes. Summary: calls api.elfa.ai only, touches nothing outside the project, writes no memory.
Behavior trace
Honeypot run · 71 s · 2 scenarios · 0 deviationsStatic findings
No findings across 3,332 lines in 8 files.
Cost to the agent
Tokens estimated at four characters each. The tool list is re-read by the agent on every turn, so it is the number to compare between servers that do the same job: ranked across the index.
Known vulnerabilities
Not checked against OSV yet; the weekly check has not reached this version.