The Agent Skills Directory

[SAFE]: The skill implements a strict human-in-the-loop model where all agent-generated content must be approved by a human before being published, significantly reducing the risk of automated errors or malicious activity.- [SAFE]: The Community Strategist agent defines an 'Account Estate' protocol which acts as a security feature. It provides clear instructions for securing corporate social accounts, including the use of multi-factor authentication (MFA), corporate-controlled recovery paths, and structured offboarding procedures to prevent unauthorized access or loss of control over brand identities.- [INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection as it ingests untrusted data from social media platforms (comments, mentions, and forum threads). Evidence includes ingestion points in social-listening-analyst.md and social-reddit-specialist.md, and capabilities to draft public-facing content. However, the risk is effectively mitigated by the mandate for human approval of all outputs and the role's focus on analysis rather than autonomous action.- [SAFE]: All external resources, including GitHub repositories and informational links, are used for documentation and methodology citations. No automated remote code execution or package installations from unknown sources are present. All referenced services like Slack, YouTube, and LinkedIn are well-known and trusted platforms.

social-media-ops