meme-pumper
Pass
Audited by Gen Agent Trust Hub on Nov 8, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests user-supplied text (such as narratives, token symbols, and milestone details) and interpolates it into markdown templates via a TypeScript CLI tool. This design creates a surface for indirect prompt injection where malicious instructions hidden in the input could be executed by the agent during subsequent tasks.\n
- Ingestion points: CLI arguments
token,narrative,emotion,fud, andmilestoneinscripts/content-generator.ts.\n - Boundary markers: Absent. User inputs are directly embedded into templates without delimiters or 'ignore' instructions.\n
- Capability inventory: The agent has access to
WebSearch,WebFetch,TodoWrite, andmcp__perplexity-ask__search.\n - Sanitization: Zod validation is used for type and enum safety but does not filter for natural language instructions or escape malicious payloads in string fields.\n- [COMMAND_EXECUTION]: The skill facilitates the execution of its own local TypeScript script (
scripts/content-generator.ts) using thenpx tsxrunner to automate marketing material generation.\n - The
SKILL.mdprovides numerous examples of command-line usage to generate raid kits, content calendars, and viral hooks.\n - The execution leverages the standard
commanderlibrary for parsing arguments.\n - This behavior is a core, legitimate function of the skill and involves execution of local, static code provided with the package.
Audit Metadata