browser-replay
Pass
Audited by Gen Agent Trust Hub on May 4, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill uses
npx -yto download and executeruvector@0.2.25and@claude-flow/cli@latestfrom the npm registry. These are utilized for managing session status and retrieving selectors from memory.- [COMMAND_EXECUTION]: Shell commands are executed via theBashtool to interface with the session management CLI (rvf status) and perform memory searches for similar selectors.- [REMOTE_CODE_EXECUTION]: The skill performs browser automation that includes executing arbitrary JavaScript viamcp__claude-flow__browser_eval, with the execution logic driven by data found in the externaltrajectory.ndjsonfile.- [PROMPT_INJECTION]: An indirect prompt injection surface is present because the skill ingests and acts upon untrusted data from an external trajectory file. - Ingestion points: Reads trajectory steps from
trajectory.ndjsonusing theReadtool. - Boundary markers: None provided; the skill is instructed to iterate through and execute each step in the trajectory.
- Capability inventory: Includes powerful browser tools (
browser_eval,browser_fill) and shell command execution viaBash. - Sanitization: There is no validation or sanitization performed on the actions or arguments loaded from the trajectory file.
Audit Metadata