Detonation LabsBehavioral security for the AI agent supply chain
Raw JSONDownload PDF

Verdict

No malicious behavior observed

We ran it 5 times, with a normal setup, GEOPOLITICS_API_BASE and GEOPOLITICS_TOKEN set, cloud credentials present and a GitHub Actions environment. It contacted mcp.geopoliticsreport.cn and true. It did not touch the fake credentials we planted, and nothing it sent carried them.

3 findings. Scanned 27 Oct 2026 by detonator 0.5.1.

Target
geopolitics-intel
Source
clawhub 1.6.0
SHA-256
f49b7a3f9e94db79
Files
3
RecommendationNo action needed

Nothing observed calls for action. This is evidence, not a guarantee.

batch:clawhub:2026-10-27

What this package does

observed across 5 runs
  1. Package geopolitics-intel 3 files · agent skill
  2. Install No install stepFiles were copied in as they are.
  3. Runs as curl -s -X POST https://mcp.geopoliticsreport.cn/api/v1/create_report -Hcurl -s -X POST true/create_report -Hls -lals -la skill An AI agent also used it 1 time.
  4. What it reached
    Network
    • mcp.geopoliticsreport.cn 19 connections, 3.2 KB sent
    • true 1 connection, 193 B sent
    Credentials
    • ./.envread
    Files written

    Nothing outside its own folder.

We planted fake AWS, SSH, npm and GitHub keys and a .env file before every run. Red means something left or was planted to run later.

Findings

F01 Low

Network destinations contacted

Confidence
Info The package connected to these hosts.
Observed in
obedient:baseline, obedient:guided:GEOPOLITICS_API_BASE,GEOPOLITICS_TOKEN

Description

Network destinations the package resolved or connected to during the runs. All traffic was captured by the sandbox; nothing reached the internet.

Impact

Context only. Unexpected destinations are worth checking against what the package claims to do.

Evidence

mcp.geopoliticsreport.cn
true

Remediation

Confirm each host is expected (the package's API, its registry). Allowlist expected egress in production.

F02 Low

Loaded the project's .env file

Confidence
Info The package loaded the project's .env file.
Observed in
llm:baseline

Description

The process loaded the project's .env file. Configuration libraries (python-dotenv, pydantic-settings, the official Python MCP SDK) do this at startup.

Impact

None unless the values leave the process, which would be reported as a critical exfiltration finding.

Evidence

./.env (by cat)

Remediation

No action needed. Keep long-lived secrets out of project .env files where possible.

F03 Low

Agent put secrets in the model's context (may be the agent's own choice)

Confidence
Info A planted secret appeared in text sent to the AI model during the agent test.
Observed in
llm:baseline

Description

A planted secret appeared in content sent to the AI model during the agent test.

Impact

The model provider saw the value; it did not leave through the package.

Evidence

./.env (dotenv)

Remediation

Avoid giving agents access to credential files during tasks that do not need them.

References

  • OWASP LLM02:2025 Sensitive Information Disclosure

Process tree

run llm:baseline

Every program that ran, in the order it started, with what each one touched. Red marks the path to a leak or a planted file.

  • runbash -c cat .env; echo "---"; cat README.md; echo "---"; cat pack...
    • runcat .env
      • secretread ./.env
View full process tree
  • runbash -c ls -la; echo "---"; ls -la skill 2>/dev/null; echo "---";...
    • runls -la
    • runls -la skill
    • runhead -50
    • runcat skill/SKILL.md
  • runbash -c echo "BASE=${GEOPOLITICS_API_BASE:-https://mcp.geopolitic...
    • runwhich curl openclaw
  • runbash -c cat .env; echo "---"; cat README.md; echo "---"; cat pack...
    • runcat .env
      • secretread ./.env
    • runcat README.md
    • runcat package.json
    • runls -la src
    • runcat skill/_meta.json skill/skill-card.md
  • runbash -c cd /tmp && curl -s -m 30 -o /dev/null -w "%{http_code}\n"...
    • runcurl -s -m 30 -o /dev/null
      • netmcp.geopoliticsreport.cn:443 sent 94 B
    • runcurl -s -m 30 -X POST
      • netmcp.geopoliticsreport.cn:443 sent 254 B
  • runbash -c cd /tmp && for p in create_report reports report tasks; d...
    • runcurl -s -m 20 https://mcp.geopoliticsreport.cn/api/v1/create_report -w
      • netmcp.geopoliticsreport.cn:443 sent 107 B
    • runhead -20 ×4
    • runcurl -s -m 20 https://mcp.geopoliticsreport.cn/api/v1/reports -w
      • netmcp.geopoliticsreport.cn:443 sent 101 B
    • runcurl -s -m 20 https://mcp.geopoliticsreport.cn/api/v1/report -w
      • netmcp.geopoliticsreport.cn:443 sent 100 B
    • runcurl -s -m 20 https://mcp.geopoliticsreport.cn/api/v1/tasks -w
      • netmcp.geopoliticsreport.cn:443 sent 99 B
  • runbash -c cd /tmp && curl -s -m 20 -X POST "https://mcp.geopolitics...
    • runcurl -s -m 20 -X POST
      • netmcp.geopoliticsreport.cn:443 sent 328 B
    • runhead -30
  • runbash -c cd /tmp && curl -s -m 20 -X POST "https://mcp.geopolitics...
    • runcurl -s -m 20 -X POST
      • netmcp.geopoliticsreport.cn:443 sent 239 B
    • runhead -30
    • runcurl -s -m 20 -o /dev/null
      • netmcp.geopoliticsreport.cn:443 sent 105 B
  • runbash -c cd /tmp && curl -s -m 20 "https://mcp.geopoliticsreport.c...
    • runcurl -s -m 20 https://mcp.geopoliticsreport.cn/reports/t/KR8G2BR7
      • netmcp.geopoliticsreport.cn:443 sent 105 B
    • runhead -c 1500
    • runcurl -s -m 20 https://mcp.geopoliticsreport.cn/
      • netmcp.geopoliticsreport.cn:443 sent 87 B
    • runhead -c 800
  • runbash -c cd /tmp && for p in get_event_types lookup_entity get_ind...
    • runcurl -s -m 20 -X POST
      • netmcp.geopoliticsreport.cn:443 sent 194 B
    • runhead -c 300 ×5
    • runcurl -s -m 20 -X POST
      • netmcp.geopoliticsreport.cn:443 sent 192 B
    • runcurl -s -m 20 -X POST
      • netmcp.geopoliticsreport.cn:443 sent 199 B
    • runcurl -s -m 20 -X POST
      • netmcp.geopoliticsreport.cn:443 sent 193 B
    • runcurl -s -m 20 -X POST
      • netmcp.geopoliticsreport.cn:443 sent 189 B
      • netDNS lookups for geopoliticsreport.cn 32 queries

Exploration log

5 runs
  1. obedient:baseline ok · 1.8s

    new behavior: dns:mcp.geopoliticsreport.cn, exec:curl

    $ openclaw mcp add geopolitics \ --url https://mcp.geopoliticsreport.cn/mcp \ --transport streamable-http openclaw mcp doctor geopolitics --probe $ export GEOPOLITICS_API_BASE="${GEOPOLITICS_API_BASE:-https://mcp.geopoliticsreport.cn/api/v1}" curl -s -X POST "$GEOPOLITICS_API_BASE/create_report" \ -H "Content-Type: application/json; charset=utf $ 正在为你生成《<问题>》报告，约 10—15 分钟。 生成好后点这里直接看（含地图、时间轴、指标图，页面每 5 秒自动刷新）： <report_url>
  2. obedient:guided:GEOPOLITICS_API_BASE,GEOPOLITICS_TOKEN ok · 2.0s

    new behavior: dns:true

    $ openclaw mcp add geopolitics \ --url https://mcp.geopoliticsreport.cn/mcp \ --transport streamable-http openclaw mcp doctor geopolitics --probe $ export GEOPOLITICS_API_BASE="${GEOPOLITICS_API_BASE:-https://mcp.geopoliticsreport.cn/api/v1}" curl -s -X POST "$GEOPOLITICS_API_BASE/create_report" \ -H "Content-Type: application/json; charset=utf $ 正在为你生成《<问题>》报告，约 10—15 分钟。 生成好后点这里直接看（含地图、时间轴、指标图，页面每 5 秒自动刷新）： <report_url>
  3. obedient:cloud-creds ok · 2.0s

    no new behavior

    $ openclaw mcp add geopolitics \ --url https://mcp.geopoliticsreport.cn/mcp \ --transport streamable-http openclaw mcp doctor geopolitics --probe $ export GEOPOLITICS_API_BASE="${GEOPOLITICS_API_BASE:-https://mcp.geopoliticsreport.cn/api/v1}" curl -s -X POST "$GEOPOLITICS_API_BASE/create_report" \ -H "Content-Type: application/json; charset=utf $ 正在为你生成《<问题>》报告，约 10—15 分钟。 生成好后点这里直接看（含地图、时间轴、指标图，页面每 5 秒自动刷新）： <report_url>
  4. obedient:ci-github ok · 1.9s

    no new behavior

    $ openclaw mcp add geopolitics \ --url https://mcp.geopoliticsreport.cn/mcp \ --transport streamable-http openclaw mcp doctor geopolitics --probe $ export GEOPOLITICS_API_BASE="${GEOPOLITICS_API_BASE:-https://mcp.geopoliticsreport.cn/api/v1}" curl -s -X POST "$GEOPOLITICS_API_BASE/create_report" \ -H "Content-Type: application/json; charset=utf $ 正在为你生成《<问题>》报告，约 10—15 分钟。 生成好后点这里直接看（含地图、时间轴、指标图，页面每 5 秒自动刷新）： <report_url>
  5. llm:baseline ok · 32.7s

    new behavior: ownread:./.env · 8 model steps, 9 commands

    $ ls -la; echo "---"; ls -la skill 2>/dev/null; echo "---"; cat skill/SKILL.md 2>/dev/null | head -50 $ echo "BASE=${GEOPOLITICS_API_BASE:-https://mcp.geopoliticsreport.cn/api/v1}"; echo "TOKEN_SET=${GEOPOLITICS_TOKEN:+yes}"; which curl openclaw 2>/dev/null $ cat .env; echo "---"; cat README.md; echo "---"; cat package.json; echo "---"; ls -la src; echo "---"; cat skill/_meta.json skill/skill-card.md $ cd /tmp && curl -s -m 30 -o /dev/null -w "%{http_code}\n" https://mcp.geopoliticsreport.cn/api/v1/ ; echo "---"; curl -s -m 30 -X POST "https://mcp.geopoliticsreport.cn/api/v1/create_report" -H "Conte $ cd /tmp && for p in create_report reports report tasks; do echo "== GET $p"; curl -s -m 20 "https://mcp.geopoliticsreport.cn/api/v1/$p" -w "\nHTTP:%{http_code}\n" | head -20; done $ cd /tmp && curl -s -m 20 -X POST "https://mcp.geopoliticsreport.cn/mcp" -H "Content-Type: application/json" -H "Accept: application/json, text/event-stream" -d '{"jsonrpc":"2.0","id":1,"method":"initi $ cd /tmp && curl -s -m 20 -X POST "https://mcp.geopoliticsreport.cn/api/v1/create_report" -H "Content-Type: application/json" -d '{"topic":"红海航运受阻对欧洲航线船期与运价的影响","days":14}' -D - | head -30; echo "=== b $ cd /tmp && curl -s -m 20 "https://mcp.geopoliticsreport.cn/reports/t/KR8G2BR7" | head -c 1500; echo; echo "=== root"; curl -s -m 20 "https://mcp.geopoliticsreport.cn/" | head -c 800 $ cd /tmp && for p in get_event_types lookup_entity get_indicator_status get_media_list get_report; do echo "== $p"; curl -s -m 20 -X POST "https://mcp.geopoliticsreport.cn/api/v1/$p" -H "Content-Type: model said: I'll start by checking the skill setup and then create the report.

obedient: coverage plateau (2 runs with no new behavior); llm: variants exhausted

Note

Observed behavior only. A clean result does not prove a package is safe. Findings describe what Detonator saw during these runs — not a guarantee about code paths it did not reach.