Skills/PyPI/madeleine-mcp

madeleine-mcp@0.2.0

CAUTION

Publisher claims: "Involuntary, cue-triggered memory for AI agents: sensory signals surface related memories without being asked." Observed in sandbox: no network, touches nothing outside the project, writes no memory.

Registry PyPI (MCP server)Publisher fabianimv · joined 0 days agoInstalls 0Last detonated 3 h ago
What would fix thisWatchDispute verdict

Is [email protected] safe to install?

KillerSkills ran it in the KillerSkills honeypot sandbox (2 scenarios) on 2026-11-06 and scored it 71/100 (Caution). It is not known to be malicious, but it has behaviour or code a security policy should decide on before install. Notable findings: Outbound HTTP to raw IP address; Collapsed <details> block carrying instructions. Summary: no network, touches nothing outside the project, writes no memory.

Behavior trace

Honeypot run · 0 s · 2 scenarios · 0 deviations
t+CategoryObserved action
0.0sprocessno install scripts in bundle
0.1sprocessdependency install exit 1
0.1sprocessexec /usr/bin/date
0.1sprocessexec /usr/bin/timeout
0.1sprocessexec .venv/bin/madeleine
0.1sclaimedno tools/list reply (exit 0)
0.1sprocessscript backend runs setup only
0.2sprocessno install scripts to re-run

Static findings

HIGH
Outbound HTTP to raw IP addresssrc/madeleine/integrations/claude_code.py:24 · Outbound HTTP to raw IP address — may bypass domain-based monitoringFix: Replace encoded or assembled strings with plain literals and remove undeclared network calls.
AST07
MEDIUM
URL using raw IP addresssrc/madeleine/integrations/claude_code.py:24 · URL using raw IP address — may bypass domain-based filteringFix: Replace encoded or assembled strings with plain literals and remove undeclared network calls.
AST07
HIGH
Collapsed <details> block carrying instructionsREADME.md:69 · Collapsed <details> block carrying instructionsFix: Remove the text that addresses the agent rather than the reader; describe behavior, do not instruct.
AST01
MEDIUM
File permission modification commandDockerfile:21 · File permission modification command — agent changes file ownership/permissionsFix: Request the narrowest permission that works; replace wildcards with the exact paths or hosts.
AST05
MEDIUM
Unicode escape sequence detectedsrc/madeleine/integrations/claude_code.py:74 · Unicode escape sequence detected — may contain obfuscated payloadsFix: Put hosts, commands and strings in plain source. Decoding at runtime looks identical to malware and costs points on its own.
AST07
MEDIUM
MCP server using network transport (SSE/HTTP)src/madeleine/mcp_server/__init__.py:32 · MCP server using network transport (SSE/HTTP) — data sent over networkFix: Namespace tool names and keep tool descriptions literal; no instructions to the agent inside them.
AST04
MEDIUM
Python subprocess execution detectedtests/test_types.py:123 · Python subprocess execution detectedFix: Build commands from fixed arguments, never from interpolated strings; prefer an API client over shelling out.
AST05
MEDIUM
MCP server using network transport (SSE/HTTP)src/madeleine/serve.py:33 · MCP server using network transport (SSE/HTTP) — data sent over networkFix: Namespace tool names and keep tool descriptions literal; no instructions to the agent inside them.
AST04
MEDIUM
MCP server using network transport (SSE/HTTP)src/madeleine/mcp_server/__init__.py:13 · MCP server using network transport (SSE/HTTP) — data sent over networkFix: Namespace tool names and keep tool descriptions literal; no instructions to the agent inside them.
AST04
MEDIUM
MCP server using network transport (SSE/HTTP)src/madeleine/mcp_server/__init__.py:31 · MCP server using network transport (SSE/HTTP) — data sent over networkFix: Namespace tool names and keep tool descriptions literal; no instructions to the agent inside them.
AST04

Cost to the agent

Instructions loaded
≈64 tok
1 file · 81 in bundle
Tool list per prompt
≈0 tok
0 tools · 0 per tool
Handshake
no reply
launch to tools/list, stdio
Undocumented tools
0
description under 20 chars · avg 0

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.

What would make this Verified

9 points to 80 · weights: behavior 45%, static 25%, prompt 15%, publisher 15%
  1. Unclaimed publisher≈ +8 on the totalClaim fabianimv: a verified publisher gets +50 on publisher trust. It is free and usually instant with GitHub sign-in.
  2. Hidden instructions≈ +6 on the totalRemove HTML comments, zero-width characters and white-on-white text that carry instructions. Everything the agent is meant to read should be visible to the person reading the file.
  3. Does not do what the description says≈ +5 on the totalIn the sandbox the skill never performed the task it claims (no call to the service it is for). Make sure the happy path runs without extra setup, or describe the real behavior.
  4. Outbound HTTP to raw IP address≈ +4 on the totalReplace encoded or assembled strings with plain literals and remove undeclared network calls.
  5. URL using raw IP address≈ +2 on the totalReplace encoded or assembled strings with plain literals and remove undeclared network calls.
  6. File permission modification command≈ +2 on the totalRequest the narrowest permission that works; replace wildcards with the exact paths or hosts.
Publish a new version to PyPI, then queue it here. Not your skill? Claim the handle to get verdict changes by email.Claim fabianimvPropose a patch to the maintainer