io.github.meridianmcp/meridian— mcp server

io.github.meridianmcp/meridian — independently scanned and version-tracked by SaferSkills.

by meridianmcp·MCP Server·github.com/meridianmcp/Meridian

Is io.github.meridianmcp/meridian safe to install?

SaferSkills independently audited io.github.meridianmcp/meridian (MCP Server) and scored it 45/100 (orange). The audit ran 55 deterministic rules across Security, Supply Chain, Maintenance, Transparency, and Community; it found 1 high-severity and 4 lower-severity findings. The full rule-by-rule trace and per-finding evidence are below. Free, methodology-open.

Score
45/100
●●●●●○○○○○
Re-scan~30s
Latest scan
ScannedJun 24, 2026 · 133d ago
Scans run0 over 90 days
Detectors55 checks · 5 categories
Findings4 warnings · 1 high
EngineSaferSkills 2b638c6
View methodology →
SaferSkills installs
This week0
This month0
All time0
CategoryWeightCategory scoreContribution
Securityprompt, exec, net, exfil, eval
35%
27
9.4 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 · 5 flagged

Securityscore 27 · 5 findings
HIGHLong base64-encoded blob hidden in the skill documentationSS-SKILL-INJECT-B64-PAYLOAD-01 · Prompt injection · extensions/meridian-latex/extension/manifest.json
HIGH — once decoded by the agent, an encoded payload has the same impact class as plain-text injection.
Why it matters

A base64 string of 128+ characters appears in a documentation file. Encoded prompt injection hides the hostile instruction in base64 — invisible to keyword filters — and relies on the agent's ability to decode it at runtime. There is no normal authoring reason to embed a multi-hundred-byte base64 blob in skill docs.

The exact value spotted
excerptextensions/meridian-latex/extension/manifest.json· json
3"name": "Meridian LaTeX — Structural Overleaf Assistant",
4"version": "0.0.1",
5"key": "MIIBIjANBgkqhkiG9w0BAQEFAAOCAQ8AMIIBCgKCAQEAuHs30J0XvRIf4qHtvD98OUK3v+V822nVnny4WsiX
… (106 chars elided on L5)
6"description": "Reads and edits Overleaf's live .tex structure (sections, tables, equations,
… (106 chars elided on L6)
7"permissions": ["activeTab", "storage", "scripting", "alarms", "tabs"],
Occurrences
1 occurrence · at L5
How to fix
Remove the encoded blob, or decode it and review what it actually contains.
  1. Decode the base64 string and confirm it is not an instruction directed at the agent.
  2. Move any legitimate binary or signature data into a dedicated file (*.sig, SIGNATURES) outside the documentation.
Framework references
OWASPLLM01ATLASAML.T0051
Trace & refs
ruleSS-SKILL-INJECT-B64-PAYLOAD-01sha25625ba19a53caf5031…rubric 365aacaView on GitHub
MEDIUMInstruction telling the agent not to ask for approvalSS-SKILL-INJECT-DONT-ASK-01 · Prompt injection · AGENTS.md×2
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
excerptAGENTS.md· markdown
232 
2331. **Start:** call `start_session(project_id, session_name)` first thing.
2342. **Execute, don't ask:** if the response has pending sprint items, claim the first
235unclaimed one (`claim_sprint_item`) and start working immediately. Do NOT ask
236"what should I work on?" when a board exists.
Occurrences
2 occurrences · first at L234, also L235
Show all 2 locations
Line
File
L234
AGENTS.md
L235
AGENTS.md
How to fix
Remove the approval-skipping instruction, or scope it narrowly to a specific safe, reversible action.
  1. Delete blanket "don't ask / no need to confirm" directives from the skill.
  2. If the skill is a genuine autonomous job, restrict the opt-out to a named non-destructive action rather than all actions.
Framework references
OWASPLLM01ATLASAML.T0051
Trace & refs
ruleSS-SKILL-INJECT-DONT-ASK-01sha256ce5204b52388d24f…rubric 365aacaView on GitHub
MEDIUMInstruction telling the agent not to ask for approvalSS-SKILL-INJECT-DONT-ASK-01 · Prompt injection · docs/meridian-docs-research-grade-investigation-2026-08-23.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
excerptdocs/meridian-docs-research-grade-investigation-2026-08-23.md· markdown
759 
760- The first integration should be a local Git-based adapter using an
761explicitly supplied Overleaf Git token. Meridian should never ask users to
762paste that token into a proposal, handoff, shared note, or capability
763manifest.
Occurrences
1 occurrence · at L761
How to fix
Remove the approval-skipping instruction, or scope it narrowly to a specific safe, reversible action.
  1. Delete blanket "don't ask / no need to confirm" directives from the skill.
  2. If the skill is a genuine autonomous job, restrict the opt-out to a named non-destructive action rather than all actions.
Framework references
OWASPLLM01ATLASAML.T0051
Trace & refs
ruleSS-SKILL-INJECT-DONT-ASK-01sha25614b7852251bf4666…rubric 365aacaView on GitHub
MEDIUMInstruction telling the agent not to ask for approvalSS-SKILL-INJECT-DONT-ASK-01 · Prompt injection · meridian/MERIDIAN.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
excerptmeridian/MERIDIAN.md· markdown
46- **`request_hitl(project_id, question)`** when you need a human decision before
47continuing. This is the **ONLY sanctioned mechanism** for human input.
48**NEVER ask the user a question in chat** (no inline "What should I do?", no
49"Should I proceed?"). **NEVER use the client's native question dialog**
50(e.g. Claude Code's AskUserQuestion). Both bypass the Meridian dashboard and
Occurrences
1 occurrence · at L48
How to fix
Remove the approval-skipping instruction, or scope it narrowly to a specific safe, reversible action.
  1. Delete blanket "don't ask / no need to confirm" directives from the skill.
  2. If the skill is a genuine autonomous job, restrict the opt-out to a named non-destructive action rather than all actions.
Framework references
OWASPLLM01ATLASAML.T0051
Trace & refs
ruleSS-SKILL-INJECT-DONT-ASK-01sha2564e0873a3c77e8cc3…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
Vendor response · right of reply
Are you the maintainer? Submit a response →

Audit the pieces. Scan the whole. Decide.

~30 seconds. Free. No account. Every finding cites a rule and a line of evidence.