pc-panel-review
Pass
Audited by Gen Agent Trust Hub on Oct 28, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The skill makes extensive use of the Bash tool to perform file operations and run reproducibility checks. It specifically instructs the agent to execute shell commands provided within a user's artifact directory.
- [EXTERNAL_DOWNLOADS]: The skill uses WebFetch and WebSearch to gather venue criteria from external sources. It also utilizes npx to run the paperlint utility, which involves downloading and executing packages from the npm registry.
- [REMOTE_CODE_EXECUTION]: The invocation of npx paperlint constitutes remote code execution as it downloads and runs a package at runtime.
- [DYNAMIC_EXECUTION]: The 'Artifact / reproducibility reviewer' role is instructed to 'actually cd into the artifact and run it'. This involves executing arbitrary, unverified code found within the user-supplied artifact folder to verify results.
- [INDIRECT_PROMPT_INJECTION]: The skill has a significant exposure to indirect prompt injection because it processes untrusted inputs from multiple sources. 1. Ingestion points: The skill reads external research papers (in .tex or PDF format), artifact code and data files, and external web content via the WebFetch tool. 2. Boundary markers: The instructions do not specify any delimiters or safety markers to isolate these untrusted inputs from the agent's internal reasoning or instructions. 3. Capability inventory: The skill possesses high-privilege capabilities, including the ability to write to the filesystem, execute shell commands via Bash, and perform network requests. 4. Sanitization: There is no mention of sanitizing or validating the content of the papers or artifact scripts before they are processed or executed.
Audit Metadata