keyword-cannibalization
Pass
Audited by Gen Agent Trust Hub on Sep 24, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands using the
fastlaneandasoCLI tools to manage and deploy App Store metadata. - Evidence includes instructions to run
fastlane deliver download_metadata,aso keywords analyze,aso keywords recommend, andfastlane deliveracross multiple repository paths. - [INDIRECT_PROMPT_INJECTION]: The skill processes app metadata and keywords retrieved from external storefronts, creating a surface for indirect prompt injection or command injection if the metadata contains malicious payloads.
- Ingestion points: App metadata (titles, subtitles, keywords) is downloaded from external storefronts via
fastlane deliver download_metadatainto thefastlane/metadata/directory. - Boundary markers: The instructions do not define delimiters or specific boundary markers to isolate the downloaded data from the agent's logic.
- Capability inventory: The skill possesses the ability to navigate directories (
cd), execute CLI tools (fastlane,aso), and perform network deployments (fastlane deliver). - Sanitization: There is no explicit requirement for the agent to sanitize or validate keywords extracted from metadata before using them as arguments in shell commands like
aso keywords analyze <conflicted_keyword>.
Audit Metadata