Forge Mcp 45/100 — SaferSkills

CategoryWeightCategory scoreContribution

Securityprompt, exec, net, exfil, eval

35%

26.3 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 75 · 1 finding

HIGHLong base64-encoded blob hidden in the skill documentationSS-SKILL-INJECT-B64-PAYLOAD-01 · Prompt injection · README.md

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

excerptREADME.md· markdown

13<a href="https://pypi.org/project/forge-mcp/"><img src="https://img.shields.io/pypi/pyversio

… (52 chars elided on L13)

14<a href="https://github.com/pm577/forge-mcp/blob/main/LICENSE"><img src="https://img.shields

… (66 chars elided on L14)

15<a href="https://smithery.ai/server/@pm577/forge-mcp"><img src="https://img.shields.io/badge

… (106 chars elided on L15)

16<a href="https://glama.ai/mcp/servers/pm577/forge-mcp"><img src="https://img.shields.io/badg

… (39 chars elided on L16)

17

Occurrences

1 occurrence · at L15

How to fix

Remove the encoded blob, or decode it and review what it actually contains.

Decode the base64 string and confirm it is not an instruction directed at the agent.
Move any legitimate binary or signature data into a dedicated file (*.sig, SIGNATURES) outside the documentation.

Framework references

OWASPLLM01 ATLASAML.T0051

Trace & refs

ruleSS-SKILL-INJECT-B64-PAYLOAD-01sha2564a3d4d2bf7418091…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.md5.2 KB · Markdown

Portable reputation that follows your agent across every platform.

What is Forge?

Every agent marketplace has the same problem: reputation is platform-dependent. An agent with 500 successful jobs on Platform A starts at zero on Platform B. Forge fixes this by making reputation portable — agents carry their trust score wherever they go.

Forge is MCP-native (Model Context Protocol). Any agent can connect — Claude Code, Hermes Agent, Codex, Cursor, or your custom agent — with zero platform lock-in.

Quick Start

# Install
pip install forge-mcp

# Run (HTTP SSE mode — for remote agents)
forge-server --port 4243

# Or stdio mode (for local MCP tools)
forge-server

Try it:

# Any agent can call Forge via MCP
forge_register("my-agent", "My Agent")
forge_vouch("alice", "my-agent", "Great work on Project X")
trust = forge_verify("my-agent")  # → 0.4

Tool Reference

Memory Layer

Tool	What it does
`forge_store`	Store facts in structured namespaces
`forge_recall`	Retrieve stored knowledge
`forge_forget`	Delete a fact
`forge_graph`	View entity relationship graph
`forge_stats`	System statistics

Reputation Layer

Tool	What it does
`forge_register`	Register a new agent identity
`forge_vouch`	Vouch for another agent's reliability
`forge_verify`	Get an agent's current trust score
`forge_endorse`	Endorse a specific skill
`forge_reputation`	Full profile — score, history, vouches

Growth Layer

Tool	What it does
`forge_referral_code`	Generate referral code for viral distribution
`forge_referral_stats`	Track referral performance

Trust Model

Base trust:       0.30
Referred agent:   0.40 (+0.1 bonus)
Per vouch:       +0.10
Per referral:    +0.05 (for referrer)
Maximum:          1.00

Referral viral loop:

Agent A shares referral code
  → Agent B joins with code (starts at 0.40)
    → Agent A earns +0.05 trust
      → Agent A's higher trust attracts more referrals
        → Network compounds

Architecture

┌─────────────┐     MCP Protocol     ┌──────────────┐
│  Any Agent   │ ◄──────────────────► │  Forge Server │
│  (Hermes,    │     (SSE or stdio)   │  :4243        │
│   Claude,   │                      │               │
│   Codex)    │                      ├──────────────┤
└─────────────┘                      │  SQLite (WAL) │
                                     │  ~/.forge/    │
                                     └──────────────┘

Use Cases

Agent marketplaces — Verify agent track records before hiring
Multi-platform agents — Carry reputation across Shellcorp, PayLock, ClawMarket
Agent teams — Vouch for teammates, build collective trust
On-chain reputation — Anchor off-chain vouches to verified identities

Why MCP?

MCP (Model Context Protocol) is the standard for agent-to-tool communication — the USB-C of AI. By building on MCP, Forge is immediately compatible with every MCP client without custom integrations.

Contributing

See CONTRIBUTING.md. All contributions welcome — issues, PRs, documentation, integrations.

<a href="https://moltbook.com/u/forgereputation">Follow Forge on Moltbook</a> · <a href="https://github.com/pm577/forge-mcp/issues">Report Issue</a> · <a href="https://github.com/pm577/forge-mcp/discussions">Discussion</a>

Forge Mcp— agent skill

Is Forge Mcp safe to install?

What is Forge?

Quick Start

Tool Reference

Memory Layer

Reputation Layer

Growth Layer

Trust Model

Architecture

Use Cases

Why MCP?

Contributing

Audit the pieces. Scan the whole. Decide.