Verdict

No malicious behavior observed

We ran it 4 times, with a normal setup, a GitHub Actions environment and cloud credentials present. It contacted pyai.com, pypi.org and 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: pyai-speak
Source: clawhub 1.0.0
SHA-256: 8495ee46ef248043
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 4 runs

Package pyai-speak 3 files · agent skill
Install No install stepFiles were copied in as they are.
Runs as ls -lased s/=.*/=<redacted>/ .envcat README.md package.jsonls -la src skill An AI agent also used it 1 time.
What it reached
Network
- pyai.com 3 connections, 222 B sent
- pypi.org 1 connection, 633 B sent
- registry.npmjs.org 5 connections, 2.0 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

pyai.com

pypi.org

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 sed)

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 npm)

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 cd "$(pwd)" && cat README.md package.json; echo "--- env ...
- runsed s/=.*/=<redacted>/ .env
  - secretread ./.env

View full process tree

runbash -c cd "$(pwd)" && pyai --version; echo "---"; pyai speak --h...
runbash -c cd "$(pwd)" && ls -la; echo "---"; pyai voices list --json
- runls -la
runbash -c cd "$(pwd)" && cat README.md package.json; echo "--- env ...
- runsed s/=.*/=<redacted>/ .env
  - secretread ./.env
- runcat README.md package.json
- runls -la src skill
- runwhich brew npm node python3 pip3
runbash -c cd "$(pwd)" && cat skill/_meta.json; echo "---"; timeout ...
- runcat skill/_meta.json
- runtimeout 15 curl -sS -o /dev/null
  - runcurl -sS -o /dev/null -w %{http_code}\n
    - netpyai.com:443 sent 74 B
- runtimeout 20 npm view @pyai/cli version
  - runnpm view @pyai/cli version
    - runnode /usr/bin/npm view @pyai/cli version
      - secretread ~/.npmrc (its own config)
      - netregistry.npmjs.org:443 sent 391 B
      - netregistry.npmjs.org:443 sent 433 B
- runtail -5
runbash -c cd "$(pwd)" && timeout 20 npm view @pyai/cli version 2>&1...
- runhead -20
- runtimeout 20 npm view @pyai/cli version
  - runnpm view @pyai/cli version
    - runnode /usr/bin/npm view @pyai/cli version
      - secretread ~/.npmrc (its own config)
      - netregistry.npmjs.org:443 sent 391 B
- runhead -c 600
- runtimeout 15 curl -sS https://pyai.com/cli
  - runcurl -sS https://pyai.com/cli
    - netpyai.com:443 sent 74 B
runbash -c cd "$(pwd)" && timeout 15 curl -sS -D - https://pyai.com/...
- runhead -30
- runtimeout 15 curl -sS -D -
  - runcurl -sS -D - https://pyai.com/cli -o
    - netpyai.com:443 sent 74 B
- runcat /tmp/pyai_cli_body.txt
- runhead -c 1000
- runwc -c /tmp/pyai_cli_body.txt
runbash -c cd "$(pwd)" && echo "=== npm registry ==="; timeout 25 np...
- runtimeout 25 npm view @pyai/cli
  - runnpm view @pyai/cli
    - runnode /usr/bin/npm view @pyai/cli
      - secretread ~/.npmrc (its own config)
      - netregistry.npmjs.org:443 sent 391 B
- runhead -20
- runtimeout 25 pip3 download pyai --no-deps
  - runpip3 download pyai --no-deps -d /tmp/pipprobe
    - runlsb_release -a
      - rungetopt --name lsb_release -o hvidrcas -l
      - runcut -c1
      - runtr [:lower:] [:upper:]
      - runcut -c2-
      - runtr [:upper:] [:lower:] ×2
    - rununame -rs
- runtail -5
runbash -c cd "$(pwd)" && echo "=== npm exit ==="; timeout 25 npm vi...
- runtimeout 25 npm view @pyai/cli version
  - runnpm view @pyai/cli version
    - runnode /usr/bin/npm view @pyai/cli version
      - secretread ~/.npmrc (its own config)
      - netpypi.org:443 sent 633 B
- runtimeout 25 pip3 download pyai --no-deps
  - runpip3 download pyai --no-deps -d /tmp/pipprobe
    - runlsb_release -a
      - rungetopt --name lsb_release -o hvidrcas -l
      - runcut -c1
      - runtr [:lower:] [:upper:]
      - runcut -c2-
      - runtr [:upper:] [:lower:] ×2
runbash -c cd "$(pwd)" && echo "=== pip ==="; timeout 25 pip3 downlo...
- runtimeout 25 pip3 download pyai --no-deps
  - runpip3 download pyai --no-deps -d /tmp/pipprobe
    - netregistry.npmjs.org:443 sent 391 B
- runcat /tmp/pip.log
- rungetent hosts pyai.com
  - netDNS lookups for npmjs.org, pyai.com, pypi.org 24 queries
- runcat /tmp/dns.log
runlsb_release -a
- rungetopt --name lsb_release -o hvidrcas -l
- runcut -c1
- runtr [:lower:] [:upper:]
- runcut -c2-
- runtr [:upper:] [:lower:] ×2
rununame -rs

Exploration log

4 runs

obedient:baseline ok · 0.4s

no new behavior

$ pyai speak --text-file "TEXT_PATH" --voice "VOICE_ID" --format wav --out "OUTPUT_PATH" --json --retries 0 --timeout 60
obedient:ci-github ok · 0.4s

no new behavior

$ pyai speak --text-file "TEXT_PATH" --voice "VOICE_ID" --format wav --out "OUTPUT_PATH" --json --retries 0 --timeout 60
obedient:cloud-creds ok · 0.4s

no new behavior

$ pyai speak --text-file "TEXT_PATH" --voice "VOICE_ID" --format wav --out "OUTPUT_PATH" --json --retries 0 --timeout 60
llm:baseline ok · 90.4s

new behavior: dns:pyai.com, dns:pypi.org, dns:registry.npmjs.org, exec:curl · 8 model steps, 9 commands

$ cd "$(pwd)" && pyai --version; echo "---"; pyai speak --help; echo "---"; pyai whoami --json $ cd "$(pwd)" && ls -la; echo "---"; pyai voices list --json $ cd "$(pwd)" && cat README.md package.json; echo "--- env keys ---"; sed 's/=.*/=<redacted>/' .env; echo "--- src ---"; ls -la src skill; echo "--- tools ---"; which brew npm node python3 pip3 curl 2>& $ cd "$(pwd)" && cat skill/_meta.json; echo "---"; timeout 15 curl -sS -o /dev/null -w "%{http_code}\n" https://pyai.com/cli 2>&1; echo "--- npm probe ---"; timeout 20 npm view @pyai/cli version 2>&1 | $ cd "$(pwd)" && timeout 20 npm view @pyai/cli version 2>&1 | head -20; echo "=== exit $? ==="; timeout 15 curl -sS https://pyai.com/cli 2>&1 | head -c 600 $ cd "$(pwd)" && timeout 15 curl -sS -D - https://pyai.com/cli -o /tmp/pyai_cli_body.txt 2>&1 | head -30; echo "=== body ==="; cat /tmp/pyai_cli_body.txt | head -c 1000; echo; echo "=== size ==="; wc -c $ cd "$(pwd)" && echo "=== npm registry ==="; timeout 25 npm view @pyai/cli 2>&1 | head -20; echo "=== pip ==="; timeout 25 pip3 download pyai --no-deps -d /tmp/pipprobe 2>&1 | tail -5; echo "=== dns == $ cd "$(pwd)" && echo "=== npm exit ==="; timeout 25 npm view @pyai/cli version; echo "exit=$?"; echo "=== pip exit ==="; timeout 25 pip3 download pyai --no-deps -d /tmp/pipprobe; echo "exit=$?"; echo " $ cd "$(pwd)" && echo "=== pip ==="; timeout 25 pip3 download pyai --no-deps -d /tmp/pipprobe > /tmp/pip.log 2>&1; echo "exit=$?"; cat /tmp/pip.log; echo "=== dns ==="; getent hosts pyai.com > /tmp/dns. model said: That's a notable finding — `https://pyai.com/cli` returned a JSON token payload rather than an install page. Let me inspect it carefully before doing anything else.

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.