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

Verdict

No malicious behavior observed

We ran it 4 times, with a normal setup, WATCH_REPORT_DEPENDENCIES set and SystemRoot set. It contacted mochilab.bz. 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
mochilab-skills
Source
clawhub 1.0.0
SHA-256
74cb3a93e6c0f0a0
Files
6
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 4 runs
  1. Package mochilab-skills 6 files · agent skill
  2. Install No install stepFiles were copied in as they are.
  3. Runs as node <skill>/scripts/mochilab.mjsls -la ./skillls -la ./skill/scriptsnode --version An AI agent also used it 1 time.
  4. What it reached
    Network
    • mochilab.bz 3 connections, 419 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

mochilab.bz

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.

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 cd . && cat README.md package.json .gitignore; echo ---; ...
    • runcat .env
      • secretread ./.env
View full process tree
  • runbash -c ls -la ./skill && ls -la ./skill/scripts 2>/dev/null; nod...
    • runls -la ./skill
    • runls -la ./skill/scripts
    • runnode --version
  • runbash -c cd . && ls -la; openclaw config get agents.defaults.works...
    • runls -la
    • runhead -20
  • runbash -c cd . && cat README.md package.json .gitignore; echo ---; ...
    • runcat .env
      • secretread ./.env
    • runcat README.md package.json .gitignore
    • runsed s/=.*/=<redacted>/
    • runls src
    • runcat skill/_meta.json skill/skill-card.md
  • runbash -c cd . && sed -n 1,120p skill/scripts/mochilab.mjs
    • runsed -n 1,120p skill/scripts/mochilab.mjs
  • runbash -c cd . && grep -n "case '\|command\|argv\|usage\|Usage" ski...
    • rungrep -n case '\|command\|argv\|usage\|Usage skill/scripts/mochilab.mjs
    • runhead -80
  • runbash -c cd . && sed -n 200,310p skill/scripts/mochilab.mjs
    • runsed -n 200,310p skill/scripts/mochilab.mjs
  • runbash -c cd . && node skill/scripts/mochilab.mjs status 2>&1 | hea...
    • runnode skill/scripts/mochilab.mjs status
    • runhead -20
    • runnode skill/scripts/mochilab.mjs help
    • runhead -30
  • runbash -c cd . && node skill/scripts/mochilab.mjs search --locale e...
    • runhead -120
    • runnode skill/scripts/mochilab.mjs search --locale en
      • netmochilab.bz:443 sent 227 B
  • runbash -c cd . && sed -n 1,60p skill/scripts/mochilab.mjs; echo ......
    • runsed -n 1,60p skill/scripts/mochilab.mjs
    • runsed -n 120,200p skill/scripts/mochilab.mjs
  • runbash -c cd . && grep -n "siteUrl\|SITE\|mochilab.bz\|catalogue is...
    • rungrep -n siteUrl\|SITE\|mochilab.bz\|catalogue is invalid\|functio... skill/scripts/mochilab.mjs
    • runhead -40
  • runbash -c cd . && curl -sS -m 20 -o /tmp/cat.json -w '%{http_code} ...
    • runcurl -sS -m 20 -o /tmp/cat.json
      • netmochilab.bz:443 sent 118 B
    • runhead -c 600 /tmp/cat.json
    • runcurl -sS -m 20 -o /dev/null
      • netmochilab.bz:443 sent 74 B
      • netDNS lookups for mochilab.bz 6 queries

Exploration log

4 runs
  1. scripts:baseline ok · 1.5s

    no new behavior

  2. scripts:guided:WATCH_REPORT_DEPENDENCIES ok · 0.9s

    no new behavior

  3. scripts:guided:SystemRoot ok · 0.7s

    no new behavior

  4. llm:baseline ok · 15.2s

    new behavior: dns:mochilab.bz, exec:curl, ownread:./.env · 8 model steps, 11 commands

    $ ls -la ./skill && ls -la ./skill/scripts 2>/dev/null; node --version $ cd . && ls -la; openclaw config get agents.defaults.workspace 2>&1 | head -20 $ cd . && cat README.md package.json .gitignore; echo ---; cat .env | sed 's/=.*/=<redacted>/'; echo ---; ls src; echo ---; cat skill/_meta.json skill/skill-card.md $ cd . && sed -n 1,120p skill/scripts/mochilab.mjs $ cd . && grep -n "case '\|command\|argv\|usage\|Usage" skill/scripts/mochilab.mjs | head -80 $ cd . && sed -n 200,310p skill/scripts/mochilab.mjs $ cd . && node skill/scripts/mochilab.mjs status 2>&1 | head -20; echo "=== help ==="; node skill/scripts/mochilab.mjs help 2>&1 | head -30 $ cd . && node skill/scripts/mochilab.mjs search --locale en 2>&1 | head -120 $ cd . && sed -n 1,60p skill/scripts/mochilab.mjs; echo ...; sed -n 120,200p skill/scripts/mochilab.mjs $ cd . && grep -n "siteUrl\|SITE\|mochilab.bz\|catalogue is invalid\|function summary\|locale" skill/scripts/mochilab.mjs | head -40 model said: Now let me check the helper's status and browse the catalogue.

scripts/scripts/mochilab.mjs: 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.