ClawAudit verdict
Skill Scanner
phy-skill-scanner
Accesses agent memory/notes AND makes external network calls
Pre-install security scanner for ClawHub skills that teaches the agent what to look for as red flags (ignore instructions, exfiltration, ~/.ssh reads); all dangerous-pattern mentions are negative examples in a security rubric, not instructions to execute.
Automated static analysis — not a human review. ClawAudit 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 ClawAudit detects co-occurrence — it does not verify that one flows into another. Read the code to confirm a live chain.
Accesses agent memory/notes AND makes external network calls — may leak personal data
LLM02 · LLM07 · ASI03 · ASI06
Accesses system credential store AND makes external network calls
LLM02 · ASI03
Accesses agent memory AND makes external network calls — may leak conversation history
LLM02 · LLM07 · ASI06
Permission integrity
network_out
agent_memory
Findings (8)
Prompt injection — tries to override agent instructions
SKILL.md · code · Ignore all previous instructions
Possible prompt injection — attempts to redefine agent identity
SKILL.md · code · You are now
Instructs covert action — may act without user awareness
SKILL.md · code · secretly
Accesses .ssh directory
SKILL.md · code · .ssh/
Accesses AWS credentials file
SKILL.md · prose · downgraded · ~/.aws/credentials
Accesses system credential store
SKILL.md · prose · downgraded · keychain
Accesses cloud provider credentials
SKILL.md · prose · downgraded · ~/.aws
Base64 encoding/decoding
SKILL.md · prose · downgraded · base64-encode
Why the tier is capped
Execution sink present in raw bytes (Hard Floor: class A). 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.
network_outcredential_storeagent_memory Thanks — recorded.