Caveat verdict
ctf-ai-ml
Reads local files AND makes external network calls
A CTF challenge reference skill for AI/ML attack techniques (adversarial examples, model extraction, prompt injection) used in legitimate security competitions; all commands are documented research techniques for solving CTF puzzles, not directed at real targets.
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.
Reads local files AND makes external network calls — the capabilities for data exfiltration co-occur (data-flow not verified)
LLM02 · LLM06 · ASI03
Permission integrity
network_out
file_read
package_install
Findings (6)
Prompt injection — tries to override agent instructions
SKILL.md · code · Ignore previous instructions
Uses eval() — can execute arbitrary code
adversarial-ml.md · code · eval(
Accesses sensitive system files
llm-attacks.md · code · /etc/passwd
Possible prompt injection — attempts to redefine agent identity
llm-attacks.md · code · You are now
os.system/popen — direct OS command execution
llm-attacks.md · code · os.popen(
HTTP request to bare IP address — common in malicious payloads
llm-attacks.md · code · http://169.254.169.254
Why the tier is capped
Execution sink present in raw bytes (Hard Floor: class 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_outnetwork_infile_read Thanks — recorded.