The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to perform research by ingesting untrusted external data, specifically through review mining (G2, Capterra, TrustRadius), competitor documentation, and changelogs. This activity creates a surface for indirect prompt injection where malicious instructions could be embedded in the content being analyzed. The skill attempts to mitigate this through its 'Evidence Discipline' rules, which require the agent to verify claims, use dated sources, and distinguish between fact and interpretation, serving as a procedural barrier against malicious or misleading input.
[SAFE]: The instructions demonstrate security-conscious design by emphasizing factual verification and avoiding unsourced claims. No evidence of obfuscation, remote code execution, persistence mechanisms, or unauthorized privilege escalation was found in the provided files. The skill manages local context files (such as .agents/product-marketing.md) in a standard manner for agent-based tasks.

competitors