ClawAudit verdict
project-trident-skill
Reads local files AND makes external network calls
Skill implements a four-tier persistent memory architecture using local SQLite, markdown files, and git backups; all data stays local and the optional vector search is self-hosted, consistent with the stated continuity purpose.
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.
Reads local files AND makes external network calls — the capabilities for data exfiltration co-occur (data-flow not verified)
LLM02 · LLM06 · ASI03
Accesses agent memory/notes AND makes external network calls — may leak personal data
LLM02 · LLM07 · ASI03 · ASI06
Accesses agent memory AND makes external network calls — may leak conversation history
LLM02 · LLM07 · ASI06
Permission integrity
network_out
file_read
agent_memory
Findings (7)
Pipe-to-shell pattern (curl | sh) — supply chain attack vector
references/platform-guide.md · code · curl -fsSL https://get.docker.com | sh
Pipe to sh — executes piped content as shell commands
references/platform-guide.md · code · | sh
Pipe to bash — executes piped content as shell commands
references/platform-guide.md · code · | bash
References sudo — requests elevated privileges
references/platform-guide.md · code · sudo
Accesses OpenClaw config/secrets directly
scripts/layer0-agent-prompt-template.md · code · ~/.openclaw/openclaw.json
References agent memory files
SKILL.md · code · MEMORY.md
Downloads executables from external URLs
references/platform-guide.md · prose · downgraded · download from https://
Why the tier is capped
Execution sink present in raw bytes (Hard Floor: class A/B/F). 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.
agent_memoryfile_readnetwork_out Thanks — recorded.