lynx-recorder-fixture

Pass

Audited by Gen Agent Trust Hub on Nov 16, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill suggests using the pnpx agent-lynx command to perform replay validation. This command fetches and executes the agent-lynx package directly from the public NPM registry.
  • [COMMAND_EXECUTION]: The skill relies on executing a local Python script (fixture_tool.py) and an external CLI tool (agent-lynx). It also suggests running a temporary local HTTP server (python3 -m http.server) to facilitate device access to fixture files during testing.
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to process and analyze external fixture files that contain structured data such as mock responses, action lists, and script metadata. These files represent an attack surface where an attacker could embed malicious instructions in text fields (e.g., returnValue or args) to influence the agent's behavior during analysis.
  • Ingestion points: The fixture_tool.py script reads data from local file paths or HTTP(S) URLs provided as arguments to its decode, summary, and roundtrip commands.
  • Boundary markers: No explicit delimiters or instructions are used to warn the agent to ignore potentially malicious embedded content within the fixture data.
  • Capability inventory: The skill possesses capabilities for file system access (read/write), network requests (via urllib.request.urlopen), and shell command execution.
  • Sanitization: While the tool performs base64 decoding and zlib decompression, it does not include sanitization or validation of the resulting JSON content for prompt injection patterns.
  • [DATA_EXFILTRATION]: The fixture_tool.py utility includes a fetch_data function capable of reading arbitrary local files or fetching data from remote URLs. While this is necessary for its primary purpose of processing fixtures, the lack of path restriction means it could be misused to access sensitive system files if the agent is directed to an incorrect path.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 16, 2026, 06:05 AM