lisa-coding-agent-parity

Pass

Audited by Gen Agent Trust Hub on Aug 17, 2026

Risk Level: SAFECOMMAND_EXECUTIONDATA_EXFILTRATIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses the Bash tool to execute several coding agent CLI tools such as claude, codex, cursor-agent, agy, and copilot to empirically verify their capabilities. It also runs a local utility script at .agents/skills/harness-parity-council/runtime-adapters.mjs.- [DATA_EXPOSURE]: The protocol involves inspecting agent-specific configuration homes including ~/.claude/, ~/.codex/, ~/.gemini/config/, and ~/.copilot/. This data access is aligned with the primary purpose of auditing installed coding agents.- [EXTERNAL_DOWNLOADS]: The skill fetches documentation and feature specifications from official websites and community repositories via the WebSearch and WebFetch tools.- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted technical data from the web (documentation, GitHub issues, and blog posts) and incorporates it into the agent's research context. This content can influence subsequent CLI queries and the final research artifact. (Ingestion points: WebSearch and WebFetch results in Step 1; Boundary markers: Absent; Capability inventory: Bash shell execution, Write and Edit file access; Sanitization: Absent).
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 17, 2026, 06:27 AM
Security Audit — agent-trust-hub — lisa-coding-agent-parity