The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and synthesize untrusted data from the web, creating a potential surface for indirect prompt injection where malicious content could influence the synthesis or the decision to launch further waves.\n
Ingestion points: Data returned by the vibe-extras:web-researcher tool in Phase 3.2.\n
Boundary markers: The prompt templates use markdown headers but do not include explicit instructions to sub-agents to ignore or sanitize commands embedded within the retrieved research results.\n
Capability inventory: The skill manages state in the /tmp directory and has the capability to launch multiple sub-agents based on the content of previous research findings.\n
Sanitization: No evidence of sanitization or filtering of the external research data was found.\n- [COMMAND_EXECUTION]: The skill uses standard shell commands for administrative purposes.\n
Evidence: Phase 1.1 requires the execution of the date command to generate timestamps for file naming and logging.

research-web