Caveat verdict
大模型token成本节约
llm-token-compressor
Accesses credentials AND makes external network calls
Token compression skill wrapping the open-source headroom project (MIT) with transparent attribution; optional data reporting to a server is disclosed upfront with user opt-in at first use, no covert exfiltration.
Automated static analysis — not a human review. Caveat flags capabilities, not confirmed intent, and can produce false positives. Disagree with this verdict? Use Dispute below.
What it does
These are capability combinations: each listed behavior occurs in the skill, but Caveat detects co-occurrence — it does not verify that one flows into another. Read the code to confirm a live chain.
Accesses credentials AND makes external network calls — potential credential theft
LLM02 · ASI03
Permission integrity
network_out
credential_access
package_install
Findings (8)
Pipe to python — executes piped content as Python code
SKILL.md · code · | python
Pipe-to-shell pattern (curl | sh) — supply chain attack vector
SKILL.md · code · curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh
Pipe to sh — executes piped content as shell commands
SKILL.md · code · | sh
Possible hardcoded credential
SKILL.md · prose · downgraded · API_KEY="opc_user_xxx
Accesses shell history/config
SKILL.md · prose · downgraded · ~/.zshrc
os.system/popen — direct OS command execution
scripts/headroom_dashboard.py · prose · downgraded · os.popen(
Python os.environ.get — reads environment variable
scripts/headroom_upload.py · prose · downgraded · os.environ.get(
Python urllib.request — network access
scripts/opc_headroom_reporter.py · prose · downgraded · urllib.request
Why the tier is capped
Execution sink present in raw bytes (Hard Floor: class A/B/D). Final tier capped at Caution — cannot be lifted by any downgrade, example-payload opt-in, or allowlist.
Permissions & capabilities
No declared permissions — minimal attack surface.
package_installnetwork_outcredential_accessnetwork_in Thanks — recorded.