Security audit
Security checks for vulnerabilities and agentic risk
This skill coherently documents credential-aware skill generation, and the flagged unsafe snippets are explicitly prohibited examples.
Review generated scripts before execution and import sealed capsules only from trusted sources: encryption does not restrict their permissions. Keep credential values out of prompts and result files. Quote the dependency requirement when installing, and confirm the required MGC version is available.
The supplied shell code performs a filesystem write and stores a plaintext credential, contradicting both the documentation-only scope and the declared credential-handling model. The '# WRONG' comment labels the command but does not prevent its execution.
As supplied as executable shell code, the chunk performs environment setup and launches a CLI, which exceeds the explicitly documentation-only purpose. Although these commands concern the named product, they are operational actions rather than specifications or templates. The snippet does not establish any credential access, encryption, sealing, or other claimed functionality.
As executable code, the statement accesses SMTP configuration instead of merely providing guidance. The 'WRONG' comment labels the pattern as undesirable but does not prevent execution. If this is only a fenced, non-executed negative example in SKILL.md, it would be consistent with the educational purpose; the supplied chunk alone does not establish that context.
Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.
# WRONG
echo "password=secret" > credentials.txt
Correct: store it in MGC and read it at runtime.
Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.
with open(TOKEN_FILE, "r") as f:
token = f.read().strip()
resp = requests.post(
f"{MGC_BASE_URL}/api/mgc/sensitive/get",
json={"info_type": info_type, "info_owner": info_owner},
headers={"X-MGC-Token": token, "Content-Type": "application/json"},
Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.
raise RuntimeError("MGC token file missing")
with open(TOKEN_FILE, "r") as f:
token = f.read().strip()
resp = requests.post(
f"{MGC_BASE_URL}/api/mgc/sensitive/get",
json={"info_type": info_type, "info_owner": info_owner},
headers={"X-MGC-Token": token, "Content-Type": "application/json"},