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

Verdict

No malicious behavior observed

We ran it 1 time, with a normal setup. It contacted tweetapi.com. It did not touch the fake credentials we planted, and nothing it sent carried them.

2 findings. Scanned 26 Oct 2026 by detonator 0.5.1.

Target
tweetapi
Source
clawhub 0.1.1
SHA-256
70d1e937707d2c1e
Files
7
RecommendationNo action needed

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

batch:clawhub:2026-10-26

What this package does

observed across 1 run
  1. Package tweetapi 7 files · agent skill
  2. Install No install stepFiles were copied in as they are.
  3. Runs as ls -Rcat references/hosted-mcp.md references/integration.md references/sources.mdls -lals -la references An AI agent also used it 1 time.
  4. What it reached
    Network
    • tweetapi.com 3 connections, 262 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
llm:baseline

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

tweetapi.com

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 sed)

Remediation

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

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 pwd; ls -la; cat README.md package.json .gitignore; echo ...
    • runsed s/=.*/=<redacted>/ .env
      • secretread ./.env
View full process tree
  • runbash -c cd ./skill && ls -R && echo "-----" && cat references/*.md
    • runls -R
    • runcat references/hosted-mcp.md references/integration.md references/sources.md
  • runbash -c cd ./skill && ls -la && ls -la references 2>/dev/null; ec...
    • runls -la
    • runls -la references
    • runcat references/hosted-mcp.md
    • runcat references/integration.md
  • runbash -c cd ./skill && ls -la . references; echo "-----"; head -60...
    • runls -la . references
    • runhead -60 references/hosted-mcp.md
  • runbash -c cd ./skill && cat _meta.json skill-card.md; ls -la agents...
    • runcat _meta.json skill-card.md
    • runls -la agents
    • runls -la
  • runbash -c cd .. && cat README.md package.json .gitignore; echo "---...
    • runcat README.md package.json .gitignore
    • runsed s/=.*/=<redacted>/ .env
    • runls -la src
    • runcat src/*
    • rungit log --oneline
    • runhead
  • runbash -c pwd; ls -la; cat README.md package.json .gitignore; echo ...
    • runsed s/=.*/=<redacted>/ .env
      • secretread ./.env
    • runls -la
    • runcat README.md package.json .gitignore
    • runls -la src
    • runcat src/index.js
    • rungit log --oneline
    • runhead
  • runbash -c cd /var/tmp/detonator-runs/622a3450e6bd45c598e5fb7c2d6d4e...
    • runcat skill/agents/openai.yaml
    • runcurl -sS -m 20 -o /tmp/llms.txt
      • nettweetapi.com:443 sent 83 B
    • runhead -40 /tmp/llms.txt
  • runbash -c cd /var/tmp/detonator-runs/622a3450e6bd45c598e5fb7c2d6d4e...
    • runcurl -sS -m 20 -D /tmp/h1
      • nettweetapi.com:443 sent 89 B
    • runhead -20 /tmp/h1
    • runhead -c 600 /tmp/ai-docs.txt
    • runcurl -sS -m 20 -o /tmp/openapi.json
      • nettweetapi.com:443 sent 90 B
      • netDNS lookups for tweetapi.com 6 queries
    • runhead -c 300 /tmp/openapi.json

Exploration log

1 run
  1. llm:baseline ok · 13.4s

    new behavior: dns:tweetapi.com, exec:curl, ownread:./.env · 8 model steps, 8 commands

    $ cd ./skill && ls -R && echo "-----" && cat references/*.md $ cd ./skill && ls -la && ls -la references 2>/dev/null; echo "-----"; cat references/hosted-mcp.md 2>/dev/null; echo "====="; cat references/integration.md 2>/dev/null $ cd ./skill && ls -la . references; echo "-----"; head -60 references/hosted-mcp.md $ cd ./skill && cat _meta.json skill-card.md; ls -la agents; echo "-----"; cd .. && ls -la $ cd .. && cat README.md package.json .gitignore; echo "--- .env keys ---"; sed 's/=.*/=<redacted>/' .env; echo "--- src ---"; ls -la src; for f in src/*; do echo "=== $f ==="; cat "$f"; done; echo "--- $ pwd; ls -la; cat README.md package.json .gitignore; echo "--- .env keys ---"; sed 's/=.*/=<redacted>/' .env; echo "--- src ---"; ls -la src; for f in src/*; do echo "=== $f ==="; cat "$f"; done; echo $ cd /var/tmp/detonator-runs/622a3450e6bd45c598e5fb7c2d6d4e97/work && cat skill/agents/openai.yaml; echo "---"; curl -sS -m 20 -o /tmp/llms.txt -w "%{http_code}\n" https://tweetapi.com/llms.txt; head -4 $ cd /var/tmp/detonator-runs/622a3450e6bd45c598e5fb7c2d6d4e97/work && curl -sS -m 20 -D /tmp/h1 -o /tmp/ai-docs.txt -w "%{http_code} %{content_type} %{size_download}\n" https://tweetapi.com/ai-docs-v2.t model said: I'll start by reading the skill's reference files to understand the available paths.

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.