Detonation LabsBehavioral security for the AI agent supply chain

Verdict

No malicious behavior observed

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

4 findings. Scanned 26 Oct 2026 by detonator 0.5.1.

Target
reelary-broll-finder
Source
clawhub 0.1.0
SHA-256
f830916e52eee8b2
Files
3
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 broll-finder 3 files · agent skill
  2. Install No install stepFiles were copied in as they are.
  3. Runs as ls -lals -la skillhead -50find . -maxdepth 3 -iname *.md An AI agent also used it 1 time.
  4. What it reached
    Network
    • registry.npmjs.org 3 connections, 1.3 KB sent
    Credentials
    • ./.envread

    ~/.npmrc read by its own tool (normal)

    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

Used the planted login with its own service (registry.npmjs.org)

Confidence
Info A planted login went to the service it belongs to, as normal authentication (npm sending its token to registry.npmjs.org while installing).
Observed in
llm:baseline

Description

A planted login was sent to the service it belongs to, only as the Authorization header of ordinary requests: for example npm sending the ~/.npmrc token to registry.npmjs.org while installing dependencies.

Impact

None in itself: this is how the package manager authenticates. Publishing, unpublishing or creating tokens with it would be reported as exfiltration.

Evidence

~/.npmrc (npm) sent only as the login header of ordinary requests (no publish)

Remediation

No action needed.

F02 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

registry.npmjs.org

Remediation

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

F03 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.

F04 Low

A tool read its own credentials (e.g. gh, aws, ssh)

Confidence
Info A tool opened its own login or config file, such as npm reading ~/.npmrc.
Observed in
llm:baseline

Description

A tool opened its own configuration or login file, such as npm reading ~/.npmrc or gh reading its hosts file.

Impact

None in itself: this is how those tools work. Listed so the report shows every credential access.

Evidence

~/.npmrc (by npx)

Remediation

No action needed.

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 README.md; echo "=== package.json"; cat package.json;...
    • 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
    • runfind . -maxdepth 3 -iname *.md
  • runbash -c which npx node reelary 2>/dev/null; echo "---"; npx -y re...
    • runwhich npx node reelary
    • runhead -40
    • runnpx -y reelary --help
      • runnode /usr/bin/npx -y reelary --help
        • secretread ~/.npmrc (its own config)
        • netregistry.npmjs.org:443 sent 433 B
        • netregistry.npmjs.org:443 sent 441 B
  • runbash -c cat README.md; echo "=== package.json"; cat package.json;...
    • runcat .env
      • secretread ./.env
    • runcat README.md
    • runcat package.json
    • runcat .gitignore
    • runls -la src
    • runcat skill/_meta.json
    • runcat skill/skill-card.md
  • runbash -c cd /tmp && npx -y reelary search "aerial city at dusk" --...
    • runnpx -y reelary search aerial city at dusk --json
      • runnode /usr/bin/npx -y reelary search aerial city at dusk
        • secretread ~/.npmrc (its own config)
        • netregistry.npmjs.org:443 sent 441 B
        • netDNS lookups for npmjs.org 6 queries
    • runhead -20
    • runnpm view reelary
      • runnode /usr/bin/npm view reelary
    • runhead -10

Exploration log

1 run
  1. llm:baseline ok · 42.9s

    new behavior: dns:registry.npmjs.org, ownread:./.env, ownread:~/.npmrc, write:~/.npm/_cacache/content-v2/sha512/5f/20/* · 3 model steps, 4 commands

    $ ls -la; echo "---"; ls -la skill 2>/dev/null; echo "---"; find . -maxdepth 3 -iname "*.md" -o -maxdepth 3 -iname "*.txt" -o -maxdepth 3 -iname "*.srt" -o -maxdepth 3 -iname "*.vtt" | head -50 $ which npx node reelary 2>/dev/null; echo "---"; npx -y reelary --help 2>&1 | head -40 $ cat README.md; echo "=== package.json"; cat package.json; echo "=== .env"; cat .env; echo "=== .gitignore"; cat .gitignore; echo "=== src"; ls -la src; echo "=== meta"; cat skill/_meta.json; echo "=== $ cd /tmp && npx -y reelary search "aerial city at dusk" --json 2>&1 | head -20; echo "--- exit: $?"; npm view reelary 2>&1 | head -10 model said: I've hit a hard blocker, so I'm stopping rather than improvising. **What I found** 1. **No narration to work from.** The project is `payments-api`, a Node/Express payments service (`src/index.js`, Stripe + Postgres). There's no script, transcript, SRT/VTT, or any video/footage files anywhere in the directory. The skill's first step is to get the narration text, and there isn't any. 2. **No foot

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.