Audit History

All 27 static findings are false positives caused by Markdown backticks, fenced blocks, field names, or ordinary market terminology. The reviewed files contain documented Longbridge CLI usage but no Ruby command substitution, system reconnaissance, prompt injection, credential access, or exfiltration intent. External service and CLI use remains an operational dependency, not evidence of malicious behavior.

All 26 static detections are false positives caused by Markdown formatting or financial terminology. No Ruby or shell backtick execution appears at the cited lines. The catalyst workflow requests private positions and account changes despite read-only, no-login metadata.

All 26 static findings were adjudicated as false positives. The flagged lines are Markdown command references, CLI help snippets, financial-field labels, or output templates, not executable code or reconnaissance behavior. No prompt injection, credential collection, destructive action, or data exfiltration intent was found.

The static backtick findings are Markdown documentation and CLI help examples, not Ruby or shell backtick execution. The low blocker findings are finance terminology or user-facing error messages, not system reconnaissance. No prompt injection, data exfiltration intent, or malicious command behavior was found in the reviewed files.

Static analysis reported many external-command and weak-crypto indicators, but the high weak-crypto findings are false positives from finance terminology and markdown formulas. The confirmed risks are legitimate external Longbridge CLI execution and network-backed MCP or WebSearch fallback for market data. No evidence found of prompt injection, credential theft, obfuscation, persistence, or destructive filesystem behavior.

All 308 static findings are false positives. The 'Ruby/shell backtick execution' matches are triple-backtick markdown code fences in documentation, not executable code. The 'Weak cryptographic algorithm' hits are financial-analysis terminology (e.g., 'block trades', 'hash' in the array sense) and ARK-methodology terms like 'Wright's Law' — not cryptography. The 'System reconnaissance' hits are market-research reconnaissance (screening/sector scans), not OS-level recon. The skill is prompt-only (no scripts/), read-only (risk_level: read_only), and requires no login. The single 'Hardcoded URL' in catalyst-radar.md:14 is an example command flag, not an exfiltration endpoint.