Caveat verdict
prompt-shield-lite
Minimal anti-injection guardrail consisting of local shell scripts (detect-injection.sh, pre-action-check.sh, pre-send-scan.sh) that analyze text locally and return exit codes; no network calls or exfiltration present.
⚠ Flagged for review — coarse, uncorroborated signal, not a confirmed exploit. Review the config yourself before installing.
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.
Findings (7)
Pipe to bash — executes piped content as shell commands
README.md · code · | bash
Prompt injection — tries to override agent instructions
README.md · code · ignore all previous instructions
Sets world-executable permissions
README.md · code · chmod 777
Accesses .ssh directory
scripts/test-v2.sh · prose · downgraded · .ssh/
Changes file ownership
README.md · prose · downgraded · chown
References tunneling service
scripts/psl-core.py · prose · downgraded · ngrok
Popular HTTP library — network access
scripts/test-v2.sh · prose · downgraded · got
Why the tier is capped
Execution sink present in raw bytes (Hard Floor: class A/B). 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.
Is this flag fair?
Thanks — recorded.